Demilie Journal of Big Data (2024) 11:5 Journal of Big Data

https://doi.org/10.1186/s40537-023-00863-9

SURVEY Open Access

Plant disease detection and classification

techniques: a comparative study
of the performances
Wubetu Barud Demilie1*   

*Correspondence:
[email protected]; Abstract
[email protected] One of the essential components of human civilization is agriculture. It helps the econ-
1
Department of Information omy in addition to supplying food. Plant leaves or crops are vulnerable to different
Technology, Wachemo diseases during agricultural cultivation. The diseases halt the growth of their respec-
University, Hossana, Ethiopia
tive species. Early and precise detection and classification of the diseases may reduce
the chance of additional damage to the plants. The detection and classification
of these diseases have become serious problems. Farmers’ typical way of predicting
and classifying plant leaf diseases can be boring and erroneous. Problems may arise
when attempting to predict the types of diseases manually. The inability to detect
and classify plant diseases quickly may result in the destruction of crop plants, resulting
in a significant decrease in products. Farmers that use computerized image process-
ing methods in their fields can reduce losses and increase productivity. Numerous
techniques have been adopted and applied in the detection and classification of plant
diseases based on images of infected leaves or crops. Researchers have made signifi-
cant progress in the detection and classification of diseases in the past by exploring
various techniques. However, improvements are required as a result of reviews, new
advancements, and discussions. The use of technology can significantly increase crop
production all around the world. Previous research has determined the robustness
of deep learning (DL) and machine learning (ML) techniques such as k-means clus-
tering (KMC), naive Bayes (NB), feed-forward neural network (FFNN), support vector
machine (SVM), k-nearest neighbor (KNN) classifier, fuzzy logic (FL), genetic algorithm
(GA), artificial neural network (ANN), convolutional neural network (CNN), and so on.
Here, from the DL and ML techniques that have been included in this particular study,
CNNs are often the favored choice for image detection and classification due to their
inherent capacity to autonomously acquire pertinent image features and grasp spatial
hierarchies. Nevertheless, the selection between conventional ML and DL hinges
upon the particular problem, the accessibility of data, and the computational capabili-
ties accessible. Accordingly, in numerous advanced image detection and classification
tasks, DL, mainly through CNNs, is preferred when ample data and computational
resources are available and show good detection and classification effects on their
datasets, but not on other datasets. Finally, in this paper, the author aims to keep
future researchers up-to-date with the performances, evaluation metrics, and results
of previously used techniques to detect and classify different forms of plant leaf or crop

© The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits
use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original
author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third
party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the mate-
rial. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or
exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://​
creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.
Demilie J ournal of Big Data (2024) 11:5 Page 2 of 43

diseases using various image-processing techniques in the artificial intelligence (AI)

field.
Keywords: Classification, Detection, DL, Image processing, ML, Plant disease

Introduction
Agricultural biodiversity is essential for providing humans with food and raw materi-
als and is an essential component of human civilization [1, 2]. The disease can occur
when pathogenic organisms such as fungi, bacteria, and nematodes; soil PH; tempera-
ture extremes; changes in the quantity of moisture and humidity in the air; and other
elements continuously harm a plant. Plant diseases can have an impact on the growth,
function, and structure of plants and crops, affecting the people that rely on them.
The majority of farmers still use manual methods to detect and classify plant ailments
because it is difficult to do so early on, and this reduces productivity. Agriculture’s pro-
ductivity is a significant economic factor. As a result, disease identification and classifi-
cation in plants are critical in agricultural industries [3]. If proper precautions are not
taken, it can have serious consequences for plants by reducing the quality, quantity, or
productivity of the corresponding products or services. Automatic disease detection and
classification recognize symptoms at an early stage, i.e., when they first appear on plant
leaves, lowering the amount of labor necessary to monitor large farms of crops.
According to [4] Plant leaf disease is a major issue in rice production, and the disease
has the potential to harm the crop, resulting in a drop in products. Farmers have a dif-
ficult time detecting and classifying plant leaf diseases. The traditional method of detect-
ing and classifying diseases by physical observation is not always reliable and may result
in a significant decrease in agricultural production [5]. Plant diseases attack the leaf
initially before infecting the entire plant, reducing production quality and quantity [6].
Recent advances in DL have resulted in numerous approaches for detecting and classify-
ing plant disease using images of infected plants [7]. Early detection and classification
of plant diseases is critical for increasing agricultural productivity [8, 9]. Plant diseases
reduce crop results by having a negative impact on the crop [10]. Plant disease identi-
fication is a major challenge in agriculture for both farmers and experts [11]. Artificial
intelligence (AI) increases crop productivity by detecting and classifying plant leaf dis-
eases early on before they spread to other plants on the farm [12]. Accurate plant disease
classification would not only increase crop results but will also provide support for vari-
ous cultivation methods [13]. Every country needs farming to meet its requirements as
well as to strengthen its economy. When crop plants are damaged by diseases, the coun-
try’s production and its economy are also affected [14, 15]. Because of data disparities,
selecting an appropriate approach for image processing is always a difficult task.
To produce good results, huge datasets necessitate advanced approaches such as CNN
and large image datasets result in increased accuracy rates [16].
Image processing is used to improve the quality of images to extract valuable informa-
tion from them; as a result of this feature, image processing techniques are used in many
areas of the medical and agricultural fields, such as color processing, remote sensing,
and pattern recognition. Image processing techniques that are acceptable, effective, and
dependable can be used to discover disease in plant leaves. Image processing can be used
in a variety of fields, including biology, agriculture, medicine, engineering, computing,
Demilie Journal of Big Data (2024) 11:5 Page 3 of 43

etc. Computerized image processing techniques are critical for detecting and classify-
ing plant diseases early before they cause widespread damage to entire crops [17, 18]. To
address this, several DL, image processing, and ML techniques were being developed to
detect and classify disease in plants using images of plant leaves. DL technologies can
help agricultural firms succeed. This research focuses on the comparative study of the
performances, evaluation metrics, and results of numerous methodologies and methods
previously used to detect and classify different forms of plant leaf diseases using image
processing approaches. Accordingly, finding a reliable technique to apply is critical to
increasing the yields of agricultural products.
The paper will have the following contributions to the scientific community.

• This paper presents an overview of recent advances in plant disease detection and
classification using ML and DL approaches. Accordingly, it provides an in-depth
review of the state-of-the-art techniques and methodologies used in the area by cov-
ering research published in the field.
• The paper shows how using ML and DL approaches improves the performance and
speed of plant disease detection and classification.
• Identifying the best DL technique for multi-class plant disease detection and classifi-
cation and optimal identification accuracy.
• The development of DL techniques for detecting and classifying numerous plant dis-
eases;
• Addressing the different labeling and class challenges in recognizing plant diseases
by recommending multi-class, multi-label DL techniques.
• The use of a new technique with different steps designed to improve plant disease
detection and classification in real-world images yields quick results and is suited for
real-time applications.

The rest of the paper is organized into different but interrelated subsections. The paper
begins by discussing plant disease identification and classification in Sect. "Plant disease
identification and classification", factors responsible for plant diseases in Sect. "Factors
responsible for plant diseases", detection and classification of plant diseases in Sect.
"Detection and classification of plant diseases", plant disease detection and classification
techniques in Sect. "Plant disease detection and classification techniques", related works
in Sect. "Related works", performance evaluation of plant disease detection and classi-
fication in Sect. "Performance evaluation of plant disease detection and classification",
results and discussions in Sect. "Results and discussions", and the conclusion and recom-
mendation in Sect. "Conclusion and recommendation".

Plant disease identification and classification

Computer vision is a subdomain of AI that allows machines to counterfeit the human
visual system and precisely draw out, inspect, and recognize real-world images in the
same way that humans do [19].
ML techniques have been used to detect and classify plant diseases, but with advance-
ments in a subset of ML, DL, this area of research appears to have considerable potential
in terms of increasing accuracy. Many developed DL architectures were used, along with
Demilie J ournal of Big Data (2024) 11:5 Page 4 of 43

various visualization techniques, to detect and classify plant disease symptoms accord-
ingly [20].
Medical diagnosis, espionage, satellite images, and agribusiness are just a few of the
rapidly increasing industries that have already shown the benefits of computer vision-
based technologies. Computer vision-enabled systems can be used in agriculture to
detect and classify plant diseases based on different features or symptoms that have been
extracted. It uses a well-defined series of steps beginning with image acquisition and
continuing with various image-processing tasks such as scaling, filtering, segmentation,
feature extraction, and selection, and finally, detection and classification are performed
using ML or DL techniques [21].

Factors responsible for plant diseases

A wide range of agricultural diseases can arise at various stages of plant development
and harm the plant’s growth, which can have a negative impact on overall crop produc-
tion [19, 22, 23]. Plant diseases are caused by a variety of conditions at various phases of
plant development [24]. As summarized in [22], crop disease-causing variables are cat-
egorized into two: biotic factors and abiotic factors. Biotic factors such as viruses, fungi,
bacteria, mites, and slugs emerge as a result of microbial infection in plants, whereas
abiotic variables such as water, temperature, irradiation, and nutritional deprivation
damage plant growth [9, 25–28]. Accordingly, some sample plant leaf images with dif-
ferent diseases from the PlantVillage dataset and different images from other datasets
showing healthy and diseased plant leaves have been included in the study [21] and dif-
ferent images from other datasets showing healthy and diseased plant leaves have been
summarized in the works of [29] and [30] accordingly. Additionally, the detail com-
puter vision-based techniques and proccsses including field crops, image acquisition,
leaf image datasets, image preprocessing (test set, training set, and validation sets), data
splitting, and performance assessment methods) for plant disease detection and classifi-
cation have been clearly indicated in the work of [21]. The details of the factors respon-
sible for plant diseases has been depicted in Fig. 1. Additionally, some sample plant leaf
images with different diseases from the PlantVillage dataset and different images from
other datasets showing healthy and diseased plant leaves have been depicted in Figs. 2. 3
respectively.

Fig. 1 Factors responsible for plant diseases [22]

Demilie Journal of Big Data (2024) 11:5 Page 5 of 43

Apple black rot Apple cedar rust Apple healthy Apple scarb Bean angular leaf spot Bean healthy

Bean rust Citrus black spot Citrus canker Citrus greening Citrus healthy Potato early blight

Potato Healthy Potato late blight Rice bacterial leaf Rice brown spot Rice leaf smut Tomato bacterial
spot
Fig. 2 Some sample plant leaf images with different diseases from the PlantVillage dataset[21]

Some sample plant leaf images with different diseases from the PlantVillage dataset
and different images from other datasets showing healthy and diseased plant leaves have
been depicted in Figs. 2, 3 respectively.
The detailed computer vision-based techniques for plant disease detection and clas-
sification have been depicted in Fig. 4.

Detection and classification of plant diseases

The detection and classification of crop diseases is an essential use of DL, ML, and com-
puter vision techniques in agriculture industries [1]. The aim is to develop algorithms
and techniques based on images of leaves or other plant features that can automatically
detect and classify agricultural plant diseases. This can help farmers assist and manage
the disease. Following a detailed and critical study of numerous recent ML and DL-based
approaches developed for plant disease detection and classification in the literature, the
author has summarised a few key challenges in crop disease detection and classification,
allowing the research community to investigate the causes that may have a significant
impact on real-time-based systems for plant identification and diagnosis. Some factors
and issues may have an impact on disease identification and classification; the majority
of them have been summarised in the studies of [19, 27, 28, 31–38].

Plant disease detection and classification techniques

Machine learning (ML) techniques or algorithms
A. The NB Technique It is a probabilistic classifier variation built on the NB classifier
idea [19, 21]. It is assumed that the patterns’ prior probabilities are known to exist and
that the class labels are assigned their posterior probabilities. In light of this premise, the
maximum likelihood values of the data that belong to a specific class label are computed
using the posterior probability. It is calculated by applying Baye’s theorem to the product
Demilie J ournal of Big Data (2024) 11:5 Page 6 of 43

A. Potato healthy B. Potato early blight C. Potato late blight D. Maize healthy

E. Maize late blight F. Maize common rust G. Maize gray leaf spot

H. Black spot I. Canker J. Greening K. Scab L. Melanose

M. Stem end root N. Anthracnose O. Chilling injury P. Green mold Q. Greasy spot
Fig. 3 Different images from other datasets showing healthy and diseased plant leaves [29, 30]

of each feature’s conditional probability. This theory works fairly well in many classifica-
tion problems, even though it usually does not hold in a real-life setting.
B. The KNN Technique It is a nonparametric, supervised ML technique commonly
applied to pattern recognition [19, 21]. It is predicated on the nearest neighbor rule,
which is applied in ML applications to classify data. This method involves training the
test pattern using the classifier, and then classifying the test pattern according to how
similar it is to each training pattern. The KNN classifier produces a class membership
value that it is a member of. The object is allocated to the most widely used class labels
among its k-nearest neighbors based on the plurality vote of its neighbors. It functions
similarly to an instance-based learning model, with locally approximated operations and
distinct computations throughout the classification process.
C. The DT Technique In supervised learning, it is a supervised classification and
regression algorithm that creates classifiers by splitting the data into multiple smaller
groups (tree structure) according to which division creates the greater disproportion
Demilie Journal of Big Data (2024) 11:5 Page 7 of 43

Fig. 4 Computer vision-based techniques for plant disease detection and classification[21]

[19, 21]. One of the often utilized attribute selection metrics that are frequently
employed as disparity measurements is the Gini index, also known as entropy. One
benefit of this method is that it may make it simple for humans to interpret the
results. If the tree could have trained without being limited by its depth, a DT may
generate very little training error. Several DT variations, including ID3, C4.5, and
CART, are widely employed in various data mining and ML applications.
D. The SVM Technique The separating hyperplane defines this supervised ML
classifier. In high-dimensional space, this technique determines the ideal hyper-
plane that maximizes the margin between the data points of the two classes [19, 21,
29]. The kernel tricks that are helpful for nonlinear classification are an attribute
of SVM. Obtaining more distinct features in the high-dimensional feature space is
highly anticipated. Several general functions, including the linear, polynomial, and
radial basis functions, can be used to transform the features to finish it. The feature
space’s dimensions could grow significantly as a result of feature transformation. As
such, it lengthens the classification process’s training period. By calculating the dot
products, it might change the features into higher proportions without changing the
feature set.
E. The RF Technique It is a collection of learning techniques for randomized DT
classifiers [19, 21]. During training, it is run by building several DTs. Based on each
classification tree’s vote, the class labels of the testing dataset are calculated. The
class labels with the highest votes by the classification trees determine the classifier’s
final result. This approach attempts to produce an uncorrelated forest of trees that
will predict performance more accurately than that of the individual tree by using
bagging and randomness of features during the building of each tree.
Demilie J ournal of Big Data (2024) 11:5 Page 8 of 43

Deep learning (DL) techniques or algorithms

A. The CNN Technique Deep feed-forward neural networks are used by the CNN to ana-
lyze multidimensional data. The CNN learns channels that are activated after it classifies
a particular highlight at some spatial positioning information [19, 21, 24, 29]. The num-
ber of epochs utilized in the implementation of various convolution filters with dimen-
sions of 2 × 2 and 3 × 3 determines their accuracy. This is contingent upon the filter’s
dimensions. Several pre-trained architectures, including VGG16, VGG19, ResNet50,
ResNet152, InceptionV3, InceptionNet, and DenseNet121, are available for use with the
CNN approach.
B. The ANN Technique A neural network is a model that mimics the information pro-
cessing capabilities of a biological system, such as the brain [19, 21, 29]. Coefficients link
artificial neurons, also known as processing elements (PEs), to create a network struc-
ture. Experience leads to the discovery of data patterns and linkages rather than their
programming. Because ANNs can comprehend complex data, they can be utilized to
extract patterns from it.

Related works
In this study, the author has analyzed the literature on various plant leaf disease detec-
tions and classifications, as well as the models/techniques that have been used.
According to [1], DL-based solutions for real-time insect detection and identification
in the soybean crop have been proposed. The performances of various transfer learn-
ing (TL) models were investigated to determine the feasibility and reliability of the pro-
posed approach for determining the insect’s identification and detection accuracy. The
proposed approach achieved 98.75%, 97%, and 97% accuracy using YoloV5, InceptionV3,
and CNN, respectively. Among these, the YoloV5 algorithm performs quite well in the
solution and can run at 53 fps, making it suitable for real-time detection. Furthermore,
a dataset of crop insects was collected and labeled by mixing images taken with various
devices. The proposed study reduced the workload of the producer, was considerably
simpler, and produced better results. The authors of [14] have proposed a system that
uses DL approaches to classify and detect plant leaf diseases. They collected the images
from the PlantVillage dataset website. They used the CNN to classify plant leaf diseases
in the suggested method. There were 15 classes, including 12 classes for diseases of
various plants that were found, such as bacteria, fungi, and so on, and three classes for
healthy leaves. As a result, they achieved high accuracy in both training and testing, with
an accuracy of 98.29% in training and 98.029% in testing for all data sets used.
In the study of [25], an effective method for recognizing and identifying rice plant dis-
ease based on the size, shape, and color of lesions in a leaf image has been presented.
The suggested model uses Otsu’s global threshold technique to perform image binariza-
tion to remove image background noise. To detect the three rice diseases, the proposed
technique based on a fully connected CNN was trained using 4000 image samples of
each diseased leaf and 4000 image samples of healthy rice leaves. The results revealed
that the proposed fully connected CNN approach was fast and effective, with an accu-
racy of 99.7% on the dataset. This accuracy far exceeded that of the existing plant dis-
ease detection and classification methods. The authors of [26] have presented a model
based on CNN to identify and classify tomato leaf disease using a public dataset and
Demilie Journal of Big Data (2024) 11:5 Page 9 of 43

complement it with images taken on the country’s farms. To avoid overfitting, genera-
tive adversarial networks were used to generate samples that were similar to the training
data. The results reveal that the proposed model performed well in the detection and
classification of diseases in tomato leaves, with an accuracy greater than 99% in both the
training and test datasets.
The authors of [29] have used the dataset “PlantVillage” to depict four bacterial infec-
tions, two viral diseases, two mold diseases, and one mite-related ailment. Images of
unaffected leaves were also shown for a total of 12 crop species. For the development
of prediction models, ML approaches such as SVMs, grey-level co-occurrence matrices
(GLCMs), and CNNs were used. AI for classification has evolved alongside the develop-
ment of the backpropagation of ANNs. Based on the real-time leaf images gathered, a
KMC operation was also performed to detect diseases. Finally, the proposed approach
achieved an overall accuracy of 99% and 98% for rice trees and apples, respectively, and
96%, 94%, 95%, and 97% for tomato trees. Multi-class classification problems, such as the
one in this study, were evaluated using precision, recall, and f-measure metrics for a set
containing only one symptom pool for each class. The authors of [39] have proposed the
use of an enhanced CNN technique to detect rice disease. DNNs have had a lot of suc-
cess with image classification tasks. In this study, they have demonstrated how DNNs
can be used for plant disease detection in the context of image classification. Finally, this
research compares existing techniques in terms of accuracy of 80%, 85%, 90%, and 95%
for TL, CNN + TL, ANN, and ECNN + GA techniques, respectively. The work in [40]
has addressed numerous ML and DL techniques. SVM, KNN, RF, LR, and CNN were
the ML approaches used in the study effort for disease prediction in plants. Then, a com-
parison of ML and DL approaches was carried out. Among the ML techniques, the RF
has the best accuracy of 97.12%; however, when compared to the DL model presented in
the study, the CNN technique has the highest accuracy of 98.43%.
The capacity to identify rice leaf disease was limited by the image backgrounds and the
conditions under which the images were acquired [41]. DL models for automated identi-
fication of rice leaf diseases suffer significantly when evaluated on independent rice leaf
disease data. The results of well-known and frequently used TL models for detecting rice
leaf disease were examined in this study. There were two methods for accomplishing this:
frozen layers and fine-tuning. The DenseNet169 findings produced an excellent testing
accuracy of 99.66%, and when the results of the fine-tuned TL models were analyzed,
Xception performed well and achieved 99.99% testing accuracy. The authors of [42] have
presented Ant Colony Optimisation with Convolution Neural Network (ACO-CNN), a
novel DL technique for disease detection and classification. ACO was used to assess the
effectiveness of disease diagnostics in plant leaves. The CNN classifier was used to sub-
tract color, texture, and plant leaf arrangement geometries from the given images. Some
of the effectiveness metrics used for analysis and providing a proposed method demon-
strate that the proposed approach outperforms previous techniques with an accuracy
rate. Aoncert measurements were utilized for the execution of these approaches. Finally,
the ACO-CNN model outperformed the C-GAN, CNN, and SGD models in terms of
accuracy, precision, recall, and f1-score. The accuracy rates of C-GAN, CNN, and SGD
were 99.6%, 99.97%, and 85%, respectively. The accuracy rate in the ACO-CNN model
was 99.98%; therefore, precision, recall, and F1-score have higher rates in the ACO-CNN
Demilie J ournal of Big Data (2024) 11:5 Page 10 of 43

technique compared to other models, and the F1-score has the highest rate compared to
other models. The authors of [43] have presented a DL model (PPLCNet) that includes
dilated convolution, a multi-level attention mechanism, and GAP layers. The model used
novel weather data augmentation to expand the sample size to enhance the generaliza-
tion and robustness of feature extraction. The feature extraction network uses saw-tooth
dilated convolution with a configurable expansion rate to extend the perceptual field of
the convolutional domain, effectively addressing the problems of insufficient data infor-
mation extraction. The lightweight CBAM attention mechanism was located in the fea-
ture extraction network’s middle layer. It was used to improve the model’s information
representation. By reducing the number and complexity of parameters computed by the
network, the GAP layer prevents overfitting of the model. The validation of the retained
test dataset reveals that the PPLC-Net model’s recognition accuracy and F1-score were
99.702% and 98.442%, respectively, and that the number of parameters and FLOPs were
15.486 M and 5.338G, respectively, which can meet the requirements of accurate and
fast recognition. Furthermore, the proposed integrated CAM visualization approach
fully validates the efficiency of the proposed model. According to the study [44], an
effective CNN model was proposed to categorize tomato leaf diseases and detect the
name of the disease affecting tomato leaves. An approach to a 2-dimensional Convolu-
tional Neural Network (2DCNN) model with 2-Max Assembling covers and completely
related layers has been proposed. The experimental results show that the model was suc-
cessful enough to detect the disease with an accuracy of 96% when compared to other
classification models such as SVM, VGG16, Inception V3, and Mobile Net CNN model.
To extract different features [45] have used model engineering (ME). To improve fea-
ture discrimination and processing speed, several SVM models were used. In the train-
ing process, the kernel parameters of the radial basis function (RBF) were computed
depending on the selected model. Six leaf image sets encompassing healthy and sick
leaves of apple, corn, cotton, grape, pepper, and rice were analyzed using PlantVillage
and UCI databases. Accordingly, the categorization procedure yielded almost 90,000
images. The findings of the experimental implementation phase reveal the potential of a
powerful model in classification activities, which would be useful for a variety of future
leaf disease diagnostic applications in the agricultural business. In terms of stability, the
dilated learning model outperforms the typical ResNet-18 design. On the test set, the
model had an average accuracy of 98.5% for leaf disease recognition models. In recogniz-
ing grape or cotton leaf diseases, a test set accuracy of 97.93% is less than the proposed
structure’s accuracy of 97.93%. The authors of [46] have presented an image segmen-
tation algorithm for the automatic detection and classification of plant leaf diseases. It
also includes an overview of various disease classification techniques that can be used to
detect plant leaf disease. The genetic algorithm was used for image segmentation, which
was vital for disease detection in plant leaf disease.
The ensemble classifiers (EC) in [47] were developed by using various approaches
to preparation, feature extraction, and classification. The performance of these multi-
ple ensemble techniques was then compared to select the best ensemble classifiers. The
suggested technique’s precision and reliability were tested in both controlled labora-
tory settings and real-world conditions using two databases, namely PlantVillage and
Taiwan tomato leaves. The top EC, which achieved 96% accuracy, was determined by
Demilie Journal of Big Data (2024) 11:5 Page 11 of 43

the consideration of shadow, brightness fluctuations, disease similarities, background

clutter, multiple leaves, and diverse textures. Here, the proposed ensemble models were
presented and linked to several DL techniques. Furthermore, the proposed solution out-
performed the most recent state-of-the-art DL technique.
The authors of [48] have proposed a hybrid DL approach for the early detection and
classification of tomato plant leaf diseases. A CNN, a convolutional attention module
(CBAM), and SVM were combined in the hybrid system. The proposed approach was
evaluated using a database of tomato leaf images. The suggested model can initially
detect nine distinct tomato diseases; however, it is not limited to this. The obtained find-
ings were highly encouraging, with an accuracy of up to 97.2%, which can be improved
by improving learning processes. Here, the proposed approach outperformed better
than the state-of-the-art DL approaches.
The proposed system was lightweight and efficient, so the farmer may install it on
any smart device with a digital camera and processing capabilities. A farmer can detect
any disease immediately with a little training, allowing them to take timely preventive
measures.
The authors of [49] have tested and presented their work on a publicly available stand-
ard database, the PlantVillage Kaggle dataset. More specifically, they obtained mAP and
accuracy values of 98.10% and 99.97%, respectively, as well as a test time of 0.23 s. Both
qualitative and quantitative results confirmed that the presented solution was robust for
plant leaf disease detection and could replace manual systems. Moreover, the suggested
method showed a low-cost solution to tomato leaf disease classification that was robust
to several image transformations, such as variations in the size, color, and orientation
of the diseased portion of the leaf. Furthermore, the framework can locate the affected
plant leaves when there are variations in blurring, noise, chrominance, and brightness.
The authors have confirmed that, based on the reported data, their technique was robust
to several tomato leaf disease classifications under varying image-capturing condi-
tions. The proposed work in [50] has employed an aggregated loss function by combin-
ing triplet and cross-entropy loss with MobileNetV2 as a basis model for the effective
classification of plant disease using small samples. For the evaluation of the proposed
study, two publicly available datasets (PlantVillage with 54,303 leaf samples and Plantdoc
with 2598 leaf samples) were used. To partition the dataset into the source and target
domains, different domain splits were examined, and a large quantity of testing was con-
ducted on the target dataset using various sample sizes.
For the analysis of the PlantVillage dataset, four domain splits were considered, and
it was found that using the proposed aggregated loss and the lightweight transfer learn-
ing (TL) model for the target domain data (K-ways, N-shot), an average improvement
in accuracy was 1.49% for split-1, 16.25% for split-2, 2.9% for split-3, and 2.1% for split-4
when compared to previous work. For the plantdoc dataset, two domain splits were eval-
uated, yielding an accuracy of around 81% with 30 samples and more than 40% with only
one sample. Using several evaluation measures such as loss functions, execution time,
model size, and model parameters, the suggested work was compared to other state-of-
the-art research works.
According to [51], a DL approach for early ginger disease detection from the leaf was
proposed through different phases. After collecting 7014 ginger images with the help
Demilie J ournal of Big Data (2024) 11:5 Page 12 of 43

of domain experts from different farms, a DL approach has been implemented. The
acquired data was subjected to various image preprocessing to design and develop a DL
model capable of detecting and classifying various scenarios. The experimental results
showed that the proposed approach was useful for detecting ginger diseases, especially
bacterial wilt. With a test accuracy of 95.2%, the proposed model can correctly detect
the given image. The results showed that the DL approach provides a fast, affordable,
and easily deployable strategy for ginger disease detection, making the model a useful
early disease detection tool. This analysis was also extended to develop a mobile app to
assist a large number of ginger farmers in developing countries.
The work in [52] has examined an alternative approach for developing a disease detec-
tion model supported by leaf classification using deep convolutional networks (DCN).
Growth in computer vision presents an opportunity to broaden and boost precision
crop protection practice, as well as expand the market for computer vision applications
in precision agriculture, a unique form of training, and thus the technique used allows
for quick and direct implementation of the system in practice. The database used in this
research work contains 77,000 images of healthy and diseased plant leaves. They were
able to train a CNN model for classifying plant diseases, determining whether they were
present or not, and then another model was trained with YOLOv7 to detect the disease,
with the trained classification model achieving an accuracy of 99.5% and the detection
model achieving mAP, precision, and recall of 65%, 59%, and 65%, respectively.
The authors of [53] have presented a DL-based CNN solution for automatically clas-
sifying and distinguishing cotton leaf diseases. There has been a lot of study done on
leaf diseases that were common in many crops, but this work offered an effective and
reliable method for identifying cotton leaf disease. The proposed method successfully
classified and detected three significant cotton leaf diseases, which were difficult to con-
trol if not detected early. The proposed model for the identification and classification of
cotton leaf diseases has used CNN, with training and testing accuracy of 100% and 90%,
respectively.
The research works of [54] have shown the detection of various diseases using hybrid
image processing and decision tree (DT) techniques. The images were from Jimma and
Zegie in Southern Ethiopia. Backpropagation artificial neural network (BPNN) and DT
approaches were used. A total of 9100 images were collected. 70% of them are used for
training, while the remaining 30% are used for testing. When a decision tree and a BPNN
with a tanh activation function were coupled the overall accuracy was 94.5%.
The authors used three different types of data sets [55] by including original images of
RGB, blending images, and a mixture of RGB images and blending images. The classifi-
cation results produced by a mixture of RGB images and blending images outperformed
others, with a Genuine Acceptance Rate (GAR) of 96.7%, followed by a percentage of
False Acceptance Rate (FAR) values of 3.3% and False Rejection Rate (FRR) values of
3.3%.
The authors of [56] have proposed a technique in the literature that reports that have
good detection accuracy but have a low reliability. This work has presented a DL-based
method for detecting tomato disease using image segmentation. The authors create leaf
masks with the VIA tool. The proposed method segments images using a customized
U-Net model and then it classifies the segmented images into ten categories using a
Demilie Journal of Big Data (2024) 11:5 Page 13 of 43

convolutional network. The accuracy of 98.12% proved that it was a promising technique
for automated tomato disease detection, which can help to improve tomato production
and reduce crop loss.
According to [57] an autonomous method for detecting and classifying coffee plant
disease becomes very crucial for better productivity. Here, the DL model was trained
using 1120 images from the Wolaita Sodo agricultural research center, and an augmen-
tation method was used to deal with data over-fitting. A total of 3360 images were used.
They have compared training from scratch and TL strategies to get the best results dur-
ing the classification of such diseases. As a result, training from scratch achieves an accu-
racy rate of 98.5%, whereas transfer-based learning achieved rates of 97.01% and 99.89%
when using Mobilnet and Resnet50, respectively. The pre-trained Resnet50 model out-
performed other approaches for the classification of images.
The authors of [58] have proposed a method for detecting tomato plant disease using a
DL-based system and image data from plant leaves. In this research, they have used the
InceptionNet model and a DL architecture based on a recently developed CNN that was
trained over 18,161 segmented and non-segmented tomato leaf images using a super-
vised learning approach to detect and recognize various tomato diseases. In this work,
two state-of-the-art semantic segmentation models, U-Net and Modified U-Net, were
used to detect and segment disease-affected regions. By 98.66%, 98.5% IoU score, and
98.73% on the dice, the Modified U-net segmentation model exceeds the standard U-net
segmentation model. Here, the InceptionNet1 achieved 99.95% accuracy for binary clas-
sification problems and 99.12% accuracy for classifying images with six segments; the
InceptionNet technique outperformed the Modified U-Net model in terms of accu-
racy. The experimental findings of their proposed method for classifying plant diseases
showed that it outperformed existing methods in the literature.
The authors of [59] have proposed a pipeline for autonomous identification of tomato
leaf diseases using three compact CNNs. The authors have used TL to extract deep fea-
tures from the CNNs’ final fully connected layer for more condensed and high-level rep-
resentation. Next, it merges elements from the three CNNs to take advantage of each
CNN structure. Following that, a hybrid feature selection approach was used to select
and build a comprehensive feature set of lower dimensions. The tomato leaf disease
identification approach has been utilized for six classifiers. The results showed that the
KNN and SVM techniques achieved the highest accuracy of 99.92% and 99.90% using
only 22 and 24 features, respectively. The proposed pipeline’s experimental findings were
also compared with existing research studies for tomato leaf disease classification, con-
firming its competitive potential.
The authors of [60] have combined features taken from input images using two recent
pre-trained CNN models, ResNet50 and MobileNet, using a deep feature concatena-
tion (DFC) technique. As a result, they have proposed MobiResNet: a neural network
that was a concatenation of the ResNet50 and MobileNet models for overall prediction
capability improvement. 5400 olive leaf images were taken from an olive grove using a
remote-controlled agricultural unmanned aerial vehicle (UAV) equipped with a cam-
era to create the dataset used in the study. The overall performance of the MobiResNet
model was 97.08%, demonstrating its superiority over ResNet50 and MobileNet, which
attained classification accuracies of 94.86% and 95.63%, respectively.
Demilie J ournal of Big Data (2024) 11:5 Page 14 of 43

The PlantDet, a robust new deep ensemble model built on InceptionResNetV2, Effi-
cientNetV2L, and Xception, has been proposed in [61]. The PlantDet not only solves
under-fitting problems, but it also nourishes performances for a scarce dataset with a
sparse number of different background image datasets. PlantDet integrates efficient
data augmentation, pre-processing, a Global Average Pooling layer, a Dropout mecha-
nism, L2 regularizes, a PRelu activation function, Batch Normalization layers, and more
Dense layers, which make the model more robust in comparison to all existing models
and assist in dealing with under-fitting and overfitting problems while maintaining high
performance.
PlantDet outperformed the previous best model for the rice leaf dataset, with 98.53%
accuracy, 98.50% precision, 98.35% recall, 98.42% F1-score, and 99.71% specificity.
PlantDet also outperformed all available base models, including several robust ensem-
ble models, on the Betel leaf dataset. Finally, Grad-CAM and Score-CAM were accom-
plished using the Xception approach to explain model performances, specifically how
DL models function for this complicated dataset. In terms of localizing the predicted
area, Score-CAM somewhat exceeded Grad-CAM +  + .
The authors of [62] have introduced a strong plant disease classification system using
a Custom CenterNet framework with DenseNet-77 as a base network. The presented
method consists of three steps. Annotations were developed in the initial phase to
determine the region of interest. Second, an enhanced CenterNet was presented, with
DenseNet-77 recommended for deep key point extraction. Finally, several plant dis-
eases were detected and classified using the one-stage detector CenterNet. They used
the PlantVillage Kaggle database to do the performance analysis, which was the standard
dataset for plant disease and challenges in terms of intensity variations, color changes,
and differences in the shapes and sizes of leaves. Both qualitative and quantitative anal-
yses confirmed that the presented method was more proficient and trustworthy than
other latest approaches in identifying and classifying plant diseases.
The goal of [63] was to assist farmers in correctly identifying early-stage diseases and
informing them about these diseases. To properly describe and categorize tomato dis-
eases, CNN was used. The entire experiment was carried out using Google Colab and
a dataset including 3000 images of tomato leaves affected by nine different diseases as
well as a healthy leaf. Firstly, the input images were first pre-processed, and the targeted
area of images was split from the original images. Secondly, the images were further pro-
cessed with different CNN model hyper-parameters. Finally, CNN extracts other fea-
tures from images like as colors, texture, and edges, among others. The results showed
that the proposed model predictions were 98.49% accurate.
The authors of [64] have proposed a tomato leaf disease classification method using
TL and feature concatenation. The authors extract features from MobileNetV2 and
NASNetMobile using pre-trained kernels (weights), then concatenate and reduce the
dimensionality of these features using kernel principal component analysis. They then
feed these features into a conventional learning algorithm. The experimental results con-
firmed the efficiency of concatenated features in improving classifier performance. The
authors have tested the three most common traditional ML classifiers, RF, SVM, and
multinomial LR, and found that multinomial LR performed the best, with an average
accuracy of 97%.
Demilie Journal of Big Data (2024) 11:5 Page 15 of 43

The authors of [65] have proposed a novel method for improving broadleaf plant cat-
egorization by combining filtered features collected by Local Binary Pattern operators
and features retrieved by plant leaf contour masks. Noise in plant images was filtered
using morphological operators of opening and closing. A testbed system was used to
collect images at four stages of growth. Local binary pattern features based on masks
were combined with filtered features and with a coefficient of k. SVM with radial basis
function kernel was used to classify crops and weeds. By investigating optimal parame-
ters, this method achieved a classification accuracy of 98.63% with 4 classes in the ‘‘bccr-
segset’’ dataset published online, compared to a previously reported method’s accuracy
of 91.85%.
Directional Local Quinary Patterns (DLQP) as a feature descriptor for plant leaf dis-
ease detection and SVM as a classifier were investigated in [66]. The DLQP as a fea-
ture descriptor was being used for the first time in horticulture for disease detection.
DLQP gives directional edge information by calculating the grey-level difference
between the reference pixel and its neighboring pixel value by involving computation
of their grey-level difference based on quinary value (− 2, − 1, 0, 1, 2) in the 0o, 45o,
90o, and 135o directions of a specified window of a plant leaf images. They have used
a research-oriented PlantVillage dataset of tomato plant (3900 leaf images) with 6 dis-
eased classes, potato plant (1,526 leaf images), and apple plant (2600 leaf images) with
3 diseased classes to examine the robustness of DLQP as a texture descriptor. Accuracy
rates of 95.6%, 96.2%, and 97.8% were obtained for the aforementioned crops, which
were higher when compared to classification on the same dataset using other standard
feature descriptors such as Local Binary Pattern (LBP) and Local Ternary Patterns (LTP).
Furthermore, the proposed method’s effectiveness was demonstrated by comparing it to
existing algorithms for plant disease phenotyping.
The authors of [67] have developed an algorithm for detecting and preventing disease
spreading to the entire crop, which results in excellent quality crop production. The
database of various leaf images was created and processed using KMC image segmen-
tation, and textural analysis of leaf images was performed using GLCM. After ranking
their attributes using an information gain algorithm, the SVM classifier was used to clas-
sify the feature-extracted images. A graphical user interface (GUI) has been developed
to portray the various stages of the image processing algorithm and to detect the two leaf
diseases.
The authors of [68] have proposed a DL-based technique for detecting tomato disease
that uses the Conditional Generative Adversarial Network (C-GAN) to generate syn-
thetic images of tomato plant leaves. Following that, a DenseNet121 model was trained
on synthetic and real images using TL to classify tomato leaf images into ten disease
categories. The proposed model has been thoroughly trained and tested using the pub-
licly available PlantVillage dataset. For tomato leaf image categorization into 5 classes, 7
classes, and 10 classes, the suggested method achieved an accuracy of 99.51%, 98.65%,
and 97.11%, respectively. Accordingly, the proposed methodology outperformed the
existing methodologies.
The authors of [69] have focused on supervised machine learning approaches such as
NB, DT, KNN, SVM, and RF for maize plant disease detection using plant images. The
aforementioned classification techniques were analyzed and compared to determine
Demilie J ournal of Big Data (2024) 11:5 Page 16 of 43

the suitable model with the highest accuracy for plant disease prediction. When com-
pared to the other classification techniques, the RF algorithm has the highest accuracy
of 79.23%.
The authors of [70] have proposed a new method of assembling TL models to detect
rice plants and categorize diseases using images. This model detects the three most
frequent rice crop diseases: brown spot, leaf smut, and bacterial leaf blight. The TL, in
general, uses pre-trained models and provides higher accuracy for image datasets. Addi-
tionally, assembling machine learning algorithms (combining two or more ML algo-
rithms) aids in reducing generalization errors and making the model more robust. EL
was becoming more popular since it minimizes generalization errors while also making
the model more robust. With an accuracy of 96.42%, they proposed a novel model that
ensembles three TL models: InceptionV3, MobileNetV2, and DenseNet121.
A newly conducted research paper microscope called Foldscope was used [71] to iden-
tify early blight disease in tomato leaves. To assess the severity of early blight disease in
tomato leaves, the study used a deep residual network 101 (ResNet101) of CNN. The
model’s dataset was trained using an open database, namely the PlantVillage dataset for
mild, moderate, and severely infected tomato leaves, as well as healthy tomato leaves.
The ResNet101 architecture’s performance was compared to that of other pre-trained
CNN architectures, including Visual Geometry Group 16 (VGG16), VGG19, Goog-
leLeNet, AlexNet, and ResNet50. The deep ResNet101 architecture achieved the highest
accuracy of 94.6% among these networks.
The authors of [72] have proposed a lightweight TL-based technique for detecting
disease in tomato leaves. It uses an efficient pre-processing method to improve clas-
sification by enhancing the leaf images with illumination correction. For effective pre-
diction, the system extracts features using a hybrid model composed of a pre-trained
MobileNetV2 architecture and a classifier network. To avoid data leaks and handle the
class imbalance issue, traditional augmentation procedures were substituted by runtime
augmentation. They trained and tested on the PlantVillage dataset. Finally, the proposed
architecture attained a 99.30% accuracy.
The authors of [73] have developed methods for detecting and classifying plant disease
using diseased plant leaves. In this study, they used four different DL models (Incep-
tionV3, InceptionResnetV2, MobileNetV2, and EfficientNetB0) to classify plant diseases
using images of healthy and diseased plant leaves. They used the standard PlantVillage
dataset with 53,407 images recorded under laboratory conditions to train and test the
classifier. They used accessible datasets to train models that included 14 different plant
species, 38 different categorical disease groups, and healthy plant leaves. Using Incep-
tionV3, InceptionResNetV2, MobileNetV2, and EfficientNetB0, the implemented mod-
els achieved disease classification accuracy rates of 98.42%, 99.11%, 97.02%, and 99.56%,
respectively, which was higher than that of traditional handcrafted-feature based tech-
niques. Finally, EfficientNetB0 achieved the maximum successful classification accuracy
of 99.56%, with significantly less training time than the InceptionV3, InceptionResNetV2,
and MobileNetV2 designs.
A DCNN technique was developed by [74] to integrate an attention mechanism that
can better adapt to the diagnosis of a range of tomato leaf diseases. The network struc-
ture that was implemented primarily consists of residual blocks and attention extraction
Demilie Journal of Big Data (2024) 11:5 Page 17 of 43

modules. The model can extract complex aspects of numerous diseases with high accu-
racy. The dataset was derived from the most widely used publicly available PlantVillage
dataset. It was also found that, when compared to another model, the SE-ResNet50 was
better suited for the identification of tomato leaf diseases. On the tomato leaf diseases
dataset, the suggested model achieved an average identification accuracy of 96.81%,
according to the experiment findings.
The authors of [75] have focused on identifying tomato plant leaf disease using image
processing techniques based on image segmentation, clustering, and open-source algo-
rithms. Using the image classifier technique that has been implemented, the compari-
son image was classed as a disease-affected or normal leaf. Other algorithms that can be
used include ANN, FL, and hybrid algorithms. All of this contributes to a reliable, safe,
and accurate leaf disease system with a focus on tomato plants’ leaves. The method was
evaluated using OpenCV, one of the most used libraries for computer vision applica-
tions. Finally, when compared to other neural networks, they attained an accuracy of
98% using the proposed technique.
The authors of [76] have used a DNN for real-time tomato plant leaf disease image
classification. They were used to develop a framework that uses a series of frameworks to
classify the leaf disease, including image acquisition, feature extraction, and the tomato
plant. The MatLab platforms were used to analyze the DNN classification performance
on tomato leaves, which consists of 1,500 images of healthy and diseased leaves, respec-
tively. The tomato leaf images were well-ranked using the DNN DL technique. The DNN
results were 86.18% more accurate than existing models.
The authors of [77] have developed a DL technique for classifying tomato plant dis-
ease by combining CNN pre-trained models and fine-tuning them. The purpose of
this work was to use multiple performance indicators to compare the performance of
AlexNet, GoogleNet, Inception V3, ResNet 18, and ResNet 50. When these models were
compared, the advantages were that CNNs do not require any tiresome pre-processing
and have a faster convergence rate and lucrative training results. The model used in this
study was capable of classifying nine diseases in tomato leaves from the healthy class,
whereas the GoogleNet model with 22 layers can achieve 99.72% tomato disease classifi-
cation using the TL training method. When compared to the other architectures, Incep-
tion V3 had the lowest performance.
The authors of [78] have proposed an advanced classification model to detect and clas-
sify tomato leaf disease. Before classification with KNN, a training dataset of 450 images
was used, and image features were retrieved using different models. The AlexNet model
had a classification accuracy of 76.1%, which was the highest when compared to other
models. In general, CNNs, such as AlexNet for feature extraction and KNN for classi-
fication, were used. Using three pre-trained CNN models, the authors of [40] used TL
and feature extraction techniques to estimate the disease severity of tomato late blight
disease.
The authors of [79] have developed a computer vision way to identify the condi-
tion by recording leaf images and recognizing the disease’s probability. A DL tech-
nique was used in this study to develop a robust judgment that spans a wide range of
leaf appearances. The MobileNet V2, a small DL architecture, has been fine-tuned to
detect three types of tomato disease. When MobileNet V2 was trained using Adagrad
Demilie J ournal of Big Data (2024) 11:5 Page 18 of 43

with a batch size of 16, it achieved the best classification performance. The system
was put through its paces on 4,671 images from the PlantVillage dataset. The findings
suggested that MobileNet V2 can detect the disease with greater than 90% accuracy.
The authors of [80] have described how to develop a hybrid of SVM and LR meth-
ods to predict powdery mildew disease in tomato plants. The SVM was used here to
minimize noise in the data before it was supplied to the LR classifier. The adaptive
sampling-based noise reduction (ANR) method was used to reduce noise using an
SVM classifier. In this study, a real-world tomato Powdery Mildew Disease (TPMD)
dataset was used to create a prediction model using the proposed strategy. Individual
SVM and LR algorithms were also employed to create the prediction models. Accord-
ing to the results, the suggested classifier outperformed SVM by 3.06% and LR by
5.35%, with an accuracy of 92.37%.
The authors of [81] have described a novel plant leaf disease identification model
based on a DCNN. The DCNN model was trained in this study using an available
dataset containing 39 different classes of plant leaves and background images. They
also employed six data augmentation techniques: image flipping, gamma correction,
noise injection, principal component analysis (PCA), color augmentation, rotation,
and scaling. Different training epochs, batch sizes, and dropouts were used to train
the proposed model. They also compared them to common TL algorithms; the new
model outperformed the others when using validation data. The proposed model had
a classification accuracy of 96.46%.
The authors of [82] have proposed a technique that outperformed the existing ML
approaches in terms of accuracy. They have presented a computer vision framework
for identifying and classifying plant diseases. The proposed technique extracts local
tri-directional patterns (LTriDP) from images of plant leaves of various types. The
classification was done using multiclass SVM, and the studies were done on a tomato
leaf dataset with five different classes. Finally, experimental findings demonstrate that
the proposed framework outperforms competing approaches based on regularly used
feature descriptors, achieving an overall accuracy of 94%.
The authors of [83] have proposed a DCNN model based on TL to identify tomato
leaf disease. In this case, the model detects disease by using real-time and stored
tomato plant images. In addition, the suggested model’s performance was evaluated
using adaptive moment estimation (Adam), stochastic gradient descent (SGD), and
RMSprop optimizers. The experimental results showed that the suggested model,
which used the TL approach, was effective in the automated categorization of tomato
leaf disease. The suggested model for the Adam optimizer, which uses a TL strategy,
yields a 99.55% accuracy, which is 0.54% higher than RMSprop and 17.78% higher
than SGD optimizers.
The authors of [84] have developed a DL-based technique to detect the disease. A
CNN-based technique was used here for disease detection and classification. There
were three convolution layers, three max-pooling layers, and two fully linked layers in
this model. The experimental findings showed that the proposed model outperformed
the pre-trained models VGG16, InceptionV3, and MobileNet. The classification accu-
racy varies from 76 to 100% depending on the class, and the suggested model’s aver-
age accuracy is 91.2% for the 9 diseases and 1 healthy class.
Demilie Journal of Big Data (2024) 11:5 Page 19 of 43

The authors of [85] have proposed many ML methods and CNN models for the iden-
tification of tomato crop disease. A simpler CNN model with eight hidden layers was
proposed. They used the publicly available PlantVillage dataset, and the proposed light-
weight model outperforms established ML approaches. While KNN provides the best
accuracy of 94.9% in a traditional ML technique, VGG16 in pre-trained models achieves
the best accuracy of 93.5%. Finally, the proposed model outperformed PlantVillage on
other datasets, with an accuracy of 98.7%.
The authors of [86] have provided an overview of various techniques for detecting and
classifying plant leaf disease. The study found that image processing approaches can give
an effective and accurate method of classifying plant leaf disease. The authors of [87]
have used GLCM, KMC, and SVM algorithms to classify diseases on images of citrus
leaves. Using their proposed strategy, the researchers achieved a classification accuracy
of 90%.
The authors of [88] have developed a method for detecting disease using image pro-
cessing techniques in wheat fields. This study has made a significant contribution to the
automation of agricultural methods and processes. The authors of [89] have developed a
computer-aided image-processing approach for detecting plant leaf diseases. To classify
cucumber leaf diseases, the authors have used KMC and SVM classifiers. The research-
ers in [90] carried out a study of five different modern DL models. With an accuracy of
86.799%, ResNet101V2 produced the best result. The authors of [91] have proposed a
model for identifying paddy leaf diseases. The authors achieved 97.5% recognition accu-
racy by combining the SVM classifier and DCNN algorithms.
The authors of [92] have proposed a model to detect groundnut leaf disease using
the KNN algorithm. According to the researcher’s perspective, plant diseases can be
caused by fungi, bacteria, viruses, and other organisms. Plant diseases can be identified
using image processing approaches based on images of affected leaves [93]. The authors
achieved an accuracy of 88.89% using SVM and KMC. Agricultural plant diseases that
impact farmlands cause a decrease in agricultural crop results volume. Early disease
detection controls the disease and prevents it from spreading throughout the farm [94].
A computerized approach for classifying soybean plants was used to develop a deci-
sion support system that provides users with advice. The categorization accuracy of the
authors was 93.79%.
To identify plant diseases, a new architecture called plant disease detection neural net-
work (PDDNN) was a successful detection system for plant diseases using a CNN model
[95].
Farmers cannot effectively identify plant disease with their naked eyes, and if the dis-
ease is not uncovered promptly, it may harm the entire farmland or even a neighbor-
ing farm. Accordingly, modern computer technology is required for efficient and early
plant leaf disease detection. The architectures of CNN were used to achieve an overall
classification accuracy of 86.00% for AlexNet and GoogleNet techniques. Every coun-
try’s economic fortune is determined by the quality and quantity of its crop production.
By classifying the disease at an early stage, production can be considerably increased.
According to [96], the technology ensures that the naked eye wastes time and resources.
The technology ensures success by allowing for faster and more accurate identifica-
tion of plant leaf disease. Cotton leaf disease was detected using a DCNN. The authors
Demilie J ournal of Big Data (2024) 11:5 Page 20 of 43

implemented their approach and confusion matrix in MATLAB to classify disease types.
Their results showed an average accuracy of 96%.
The authors of [97] have proposed using image processing and artificial neural
approaches to identify disease in brinjal leaf plants. For segmentation, they used the
KMC algorithm, and for classification, they used the ANN.
The authors of [98] have done a study that used DL to detect tomato leaf diseases. To
detect tomato plant disease, they used an algorithm that ran in real-time. The authors of
[99] have used CNN to investigate tomato leaf disease. The authors attained a 97% clas-
sification accuracy.
Monitoring large farms with automated approaches can be useful to farmers because
it reduces the tedious work of close observation by experts. In classifying plant leaf dis-
ease, automatic systems used image segmentation algorithms [100]. They have used
MATLAB for experimentation purposes.
The authors of [101] have proposed a new image identification system based on sev-
eral linear regressions and an enhanced histogram segmentation algorithm. Their
method worked well since the appropriate threshold could be determined automatically.
It was important to use digital image processing for vision-based automated prediction
of crop plant diseases [102]. The KMC algorithm was used for color segmentation, and
the GLCM technique was used for classification. Here, good results and promising per-
formances were achieved. The authors proposed that their technique reduces detection
time and labor costs. Using computerized imaging technologies, early detection and
identification of plant leaf disease can aid in enhancing agricultural production [103].
The authors examined various ways of detecting leaf plant diseases used by previous
researches. Their research provides a critical analysis of various image classification algo-
rithms. To identify crop plant diseases, many disease recognition approaches have been
used. Different strategies have been used to increase disease classification recognition
rates [104]. When it comes to accurately identifying plant diseases, farmers have lim-
ited knowledge. According to recent studies, computerized image classification systems
based on CNN structure can aid in addressing the issue of erroneous identifications.
According to [105] datasets for rice plant disease are not easily available. The authors
developed their small dataset and achieved a 92.46% accuracy. The authors in [106] have
written research that highlights the importance of using image processing techniques
to detect plant diseases using images of infected leaves. The researchers developed a
software system solution for automatically detecting and classifying images. They have
discussed the steps involved in the classifying process, from loading the images to iden-
tifying the disease.
Farmers lack a full understanding of leaf plant diseases and must be guided on how to
apply computerized image processing methods [107]. The researchers proposed a solu-
tion that works with Android applications. The model provides an efficient method for
detecting and classifying plant diseases. The CNN was used by the proposed system to
classify plant leaf diseases. Accordingly, DL-based disease identification in plants and
pests poses a variety of issues in the future. However, feasible answers and general sug-
gestions can be obtained through the practical implementation of different plant leaf dis-
ease detection technologies. Adopting technology is required to address the challenges
of recognizing and analyzing plant diseases [108]. The authors proposed a system for
Demilie Journal of Big Data (2024) 11:5 Page 21 of 43

processing and classifying potato leaf disease images that used various Classifier algo-
rithms to achieve 97.00% classification accuracy.
The authors of [109] have developed a methodology for developing a CNN-based
strategy for detecting and identifying plant leaf disease. The CNN and DNN can be
used to swiftly and effectively identify crop diseases and their symptoms. The authors
of [110] have designed a system that classified guava leaf diseases using a DCNN tech-
nique. Their solution achieved 98.74% accuracy by using the AlexNet strategy based on
the DCNN model. The authors of [111] have used Inception v3 to develop such an iden-
tification model. Three classification metrics were used to evaluate the proposed model.
The authors reviewed the identification of plant diseases. They also established that the
CNN technique had a higher capacity to produce accurate results.
According to [112],14 plant types and 26 diseases were used to develop a CNN
approach for detecting plant diseases. Using AlexNet and GoogleNet, the model
obtained a testing accuracy of 99.35%. The authors of [113] have used ANN to develop a
model with an 80% classification accuracy on cotton leaf. ANN has been used to classify
plant leaf diseases by different researchers. The authors of [114] have carried out a study
that used ANN on 169 images. Their approach had a 92% accuracy rate.
Another comparative study was carried out by [115]. Here, various neural networks
that have previously been used by various researchers have been investigated and
reviewed. The authors established that the most widely used neural networks were ANN
and CNN. DL algorithms were accurate in classifying plant diseases. The authors of
[116] used CNN to identify diseases in various crop plants with a 98% accuracy. The
authors of [117] have used MATLAB to perform research on the automatic classification
of plant leaf diseases. The authors experimented on beans and tea leaves, achieving accu-
racy rates of 98.2% and 98.4%, respectively.
The authors of [118] have conducted a review on the identification of plant leaf dis-
eases using digital image processing approaches. The efficiency of various image pro-
cessing algorithms has been highlighted. According to the author’s point of view, most
researchers prefer neural networks to other methods. The authors of [119] have studied
groundnut leaf disease using the backpropagation approach to classify four types of dis-
eases. Their work was 97% accurate. The authors of [120] have conducted an extensive
review of recent work on plant leaf disease classification using digital image process-
ing technologies. The authors demonstrate that DL methods outperformed conventional
approaches when large amounts of data are available.
When it comes to plant leaf disease, neural network classifiers produce greater out-
put and efficiency [121]. The authors have used multi-layer perceptron (MLP) and Radial
Basis Function (RBF) to study the wall of Phyllanthus Elegans. Their approaches pro-
vided 90.15% and 98.85% accuracy, respectively. The authors of [122] have experimented
with four different diseases using cotton and soybean leaves: angular leaf spot, bacte-
rial pustule, bacterial gummosis, and bacterial blight. They were achieved 83%, 80%,
80%, and 70%, respectively. A study on plant leaf disease recognition using CNN and a
Bayesian algorithm was performed by [123]. The authors achieved an overall accuracy of
98.9% by using over 20,000 images of potato, tomato, and pepper bells.
The authors of [124] have investigated four grape plant diseases: leaf blight, black rot,
stable, and black measles. The authors achieved a total accuracy of 98.7% by combining
Demilie J ournal of Big Data (2024) 11:5 Page 22 of 43

hybrid CNN with feature reduction. Automatic identification of plant leaf disease using
plant leaves was critical in farming. The authors [125] attained a 96% overall accuracy.
Using pattern recognition, researchers in [126] reached an accuracy of 90% in the clas-
sification of apple leaf disease.
The authors of [127] have proposed an automated strategy for identifying paddy leaf
diseases using ML and evolutionary techniques. The model can be used on the Android
platform to quickly identify diseases.
According to [128] a customized CNN-based maize plant disease identification model
was presented, along with various preprocessing techniques, including contrast-limiting
adaptive histogram equalization (CLAHE) on each RGB (Red, Green, and Blue) channel,
log transformation, and RGB to HSV (Hue, Saturation, and Variance) image conversion.
These trained models were compared to CNN and SVM models that were trained with-
out any preprocessing procedures. The experiments were carried out using the Plant-
Village maize crop dataset to determine the effectiveness of the proposed effort. The
proposed work had a maximum accuracy of 99.76%. The authors of [129] have achieved
an overall accuracy of 98.39% when using GA to classify and analyze unhealthy grape
plant leaves. A survey was conducted in [130] to examine several image classification
methods that can be used to identify diseases of plant leaves. The authors of [131] have
researched different DL approaches used in the classification of plant leaf diseases. The
authors used the Weka tool to present a thorough examination of numerous techniques.
The authors of [132] have used the NB classification method to develop a model
that can identify diseased sections of plant leaves with an overall accuracy of 97%. The
authors of [133] have used the NB algorithm to classify plant leaves. The authors fed
the shape and texture of the leaves into the classifier. The CNN model outperformed
other classifiers such as KNN, RF, SVM, and DT [134]. The authors achieved 99.58% and
97.66% accuracy in identifying paddy and potato leaf diseases, respectively. Using CNN,
the authors of [135] have developed a system that provides a report on detecting banana
leaf disease. The authors of [136] have proposed three CNN models to identify guava
leaf disease. Here, the third model produced an overall accuracy of 95.61%.
The authors of [137] have developed an ML-based technique for detecting and identi-
fying potato leaf disease. The model’s overall accuracy was 95.99%. According to [138], a
model for classifying bean leaf diseases was developed. The authors achieved 97% train-
ing accuracy and 92% testing accuracy. The authors of [139] have proposed a system for
detecting maize leaf disease using improved DCNN. Using DCNN, the authors of [140]
developed a method for detecting and classifying purple blotch disease in onions. The
accuracy of the model was 85.47%. Several DL models for detecting and identifying plant
leaf diseases have been developed. DL algorithms that offer more accuracy make this
subject of research interesting [20]. ML and DL technologies have been used in agri-
culture to increase the product while addressing difficulties [141]. To classify plant leaf
diseases, the authors of [142] have presented an attention-based depth-wise neural net-
work with Bayesian optimization (ADSNN-BO). An accuracy of 94.65% was obtained.
Using the RF technique, researchers have proposed a model that can detect and clas-
sify maize disease. An automatic method for the classification of plant diseases based on
the RF algorithm was developed by [143]. The proposed model had a 95% accuracy rate.
The authors of [144] developed a hybrid learning model for plant leaf classification. To
Demilie Journal of Big Data (2024) 11:5 Page 23 of 43

identify plant leaf diseases, the model used KMC and CNN methods. The model outper-
formed SVM and CNN with an accuracy of 92.6%.
The authors of [145] have used image SVM to identify sugarcane leaf disease and
achieved a classification accuracy of 95%. The authors of [97] have used ANN and KNN
approaches to develop a model to identify and classify cassava leaf disease. The model in
[146] achieved a classification accuracy of more than 90%.
In the work of [147] the DL and ML techniques were proposed and used to solve
binary and multiclass subcategories of plant diseases on four different datasets. The pro-
posed approach produces improved results, with 99.4% accuracy and 99.9% sensitivity
for binary classes, and 99.2% accuracy for multiclass. In terms of f-measure, recall, MCC
(Matthews correlation coefficient), specificity, and sensitivity, the proposed framework
outperformed previous approaches such as LibSVM, SMO (sequential minimal optimi-
zation), and DL with activation functions softmax and soft sign.
The authors of [148] have focused on the issue and present a DL-based technique for
disease detection and classification in maize crops. Furthermore, the developed tech-
nique returns segmented images of affected leaves, allowing them to follow the disease
spots on each leaf. A dataset of three maize crop diseases blight, sugarcane mosaic virus,
and leaf spot was collected from the University Research Farm Koont, PMAS-AAUR,
during different growth stages and under different weather conditions. The data was
used to train different models for prediction, such as YOLOv3-tiny, YOLOv4, YOLOv5s,
YOLOv7s, and YOLOv8n, with stated prediction accuracy of 69.40%, 97.50%, 88.23%,
93.30%, and 99.04%. The results showed that the YOLOv8n model outperformed the
other models in terms of prediction accuracy. With a higher confidence score, this model
produced good results in properly localizing the afflicted portion of the leaf. Further-
more, the high-accuracy models have been implemented in a smartphone application to
give end users real-time disease detection in a matter of seconds.
The authors of [149] have proposed an automated DL with a wavelet neural network
(ADLWNN) model that focuses on the effective detection and classification of rice
plants. To extract features from the input rice plant images, the proposed ADLWNN
model principally employed a CNN model. Furthermore, the manta ray optimization
algorithm (MRFO) was used as a hyperparameter optimizer. Furthermore, the WNN
model was used for the accurate detection and classification of rice plant images. The
simulation analysis of the ADLWNN model was tested using a set of rice plant images,
and the findings showed that the ADLWNN model outperformed other methodologies
by 98.17%.
The authors of [150] have provided a fully automated coffee leaf disease detection
framework based on a modified color process in which the syndrome is self-clustered
using an extended Gaussian kernel density estimate and the likelihood of the nearest
common neighborhood. Accordingly, the proposed extended Gaussian kernel, which
joins nearby lesions in one symptom cluster without the necessity for any influenc-
ing set that guides towards the correct cluster, yielded the best probabilities. Clusters
are provided with the same priority as a kernel density estimation (KDE) with the
ResNet50 classifier, reducing misclassification with an accuracy of up to 98%. To iden-
tify plant diseases, residual attention learning has been proposed. The work makes
use of four different datasets [151] the residual attention network (Res-ATTEN) is
Demilie J ournal of Big Data (2024) 11:5 Page 24 of 43

used to compute attention-aware characteristics in the system. The proposed tech-

nique has reached state-of-the-art performance, with 99% for apples and rice, 94% for
corn, and 97% for grapes.
The DeepPlantNet DL-based architecture has been proposed, and the study was
efficient, unique, and lightweight for predicting and categorizing plant leaf diseases
[152]. The DeepPlantNet model suggested consists of 28 learned layers, 25 convolu-
tional layers (ConV), and three fully connected (FC) layers. It was a unique plant dis-
ease classification system that used Leaky RelU (LReLU), batch normalization (BN),
fire modules, and a mix of 33 and 11 filters. The proposed DeepPlantNet model can
classify plant disease photos into a variety of categories. The proposed method divides
plant diseases into the ten categories listed below: peach_bacterial_spot (PBS), pep-
per_bell_bacterial_spot (PBBS), potato_early_blight (PEB), squash_powdery_mildew
(SPM), strawberry_leaf_scorch (SLS), bacterial tomato spot (TBS), and maize com-
mon rust (MCR). In the case of eight-class and three-class classification schemes, the
proposed framework attained average accuracy of 98.49 and 99.85, respectively. The
authors of [153] have proposed a PDD framework (i.e., PDD-Net) based on CNN,
which leverages data augmentation techniques and includes multilevel and multiscale
features to produce a class and scale-invariant architecture. The flatten-t swish (FTS)
activation function was used to prevent gradient disappearing and explosion difficul-
ties during PDD-Net training, while the focused loss function was used to alleviate
the impact of class imbalance. On the PlantVillage dataset, the PDD-Net technique
beats baseline models, attaining an average precision of 92.06%, recall of 92.71%,
F1 score of 92.36%, and accuracy of 93.79%. On the cassava leaf disease dataset, it
achieves an average precision of 86.41%, recall of 85.77%, F1 score of 86.02%, and
accuracy of 86.98%. These findings indicate PDD-Net’s efficiency and resilience in
plant disease diagnosis.
Early detection of plant diseases is critical because they inflict social, ecological,
and economic costs [154]. In this study, four well-known CNN models (MobileNet-
V3Small, EfficientNetV2L, InceptionV3, and MobileNetV2) are used in addition to
the suggested two novel CNN models. The experimental results show that the sug-
gested ensemble models stand out for their quick training and testing times, as well as
their excellent classification performances with a 99.60% accuracy.
The authors of [155] have proposed a DNSVM classification technique that com-
bines DenseNet-201 with SVM to identify plant leaf diseases. The PlantVillage dataset
had good variations, color differences, differences in orientation, and leaf size. Sugar-
cane plant leaves were employed for the suggested model’s performance examination,
and it achieved 97.78% classification accuracy above the previous DenseNet-121-
based classifier model (94%).
The authors of [156] have suggested two DL techniques for palm leaf disease clas-
sification: residual network (ResNet) and inception ResNet TL. The models enable
the training of hundreds of layers while achieving excellent performance. Given the
value of their excellent representation ability, the performance of image classifica-
tion using ResNet has improved, such as in the disease classification of plant leaves.
Both approaches addressed issues such as brightness and backdrop variance, image
scale variation, and inter-class similarity. The date palm dataset, which included 2,631
Demilie Journal of Big Data (2024) 11:5 Page 25 of 43

colored images of varying sizes, was utilized to train and test the models. Using cer-
tain well-known measures, the suggested models surpassed many prior studies in the
field in both original and augmented datasets, achieving an accuracy of 99.62% and
100%, respectively.
Table 1 summarizes the DL and ML-based implemented plant leaf and/or crop species
disease detection and classification research works including the type (s) of plant, used
techniques/models/algorithms, and accuracies.

Performance evaluation of plant disease detection and classification

Since agriculture is a significant part of the population around the globe, a comprehen-
sive research study on several types of DL and ML approaches for different plant leaf
and/or crop disease detection and classification has been implemented. Following that,
different classification techniques in the DL and ML approaches may be utilized for
plant disease detection and classification, to assist farmers with automatic disease detec-
tion and classification of all types of disease in the crop that were to be identified [162].
The distribution of relevant papers over the years has been depicted in Fig. 5. The Fig-
ure shows an increase in plant leaf and/or crop disease detection and classification over
the years.
Accurately detecting and classifying different plant disease occurrence at an early
stage is critical for crop quality and result through the selection of appropriate treat-
ments [150, 163]. The detection of such diseases on a large scale in a timely and accu-
rate way is prone to human error [164, 165]. Accordingly, ML and DL techniques offer
possibilities for developing automated models that can detect such diseases promptly.
However, disease identification and classification necessitate specialized knowledge and
extensive experience in plant pathology. Thus, by developing an early detection and clas-
sification system for disease, an automated system for disease detection in crops will
play an essential role in agricultural industries. Automatic plant disease detection and
classification is an interesting study area since it may be useful for monitoring enormous
fields and, hence, the automatic identification and classification of disease by symptoms
in different plant sections [166]. Image dataset size, class numbers, preprocessing tech-
niques, classification approaches, performance analysis, and so on are examples of these
measurements. The previous decade’s research was thoroughly reviewed, including
studies on several plant diseases, and an investigation of the important features was pro-
vided. This work contributes by comparing the automated detection and classification of
various plant diseases through the use of image processing, DL, ML, and meta-heuristic
optimization techniques. The study gives information on many methods used to identify
diseases in various plants. The study’s various elements include the type of segmenta-
tion, dividing technology, extracted features, dataset size and year of publication, disease
category, methodologies, detection accuracy, classification, and constraints accordingly.

Evaluation metrics
The application of DL approaches to detect and classify plant diseases can avoid the dis-
advantages of artificial selection of disease spot features, making plant disease feature
extraction more objective, and increasing research performances and technology trans-
formation speed [167]. The researchers in the literature have measured their proposed
Demilie J ournal of Big Data (2024) 11:5 Page 26 of 43

Table 1 Summary of the related works using the DL and ML techniques

Author(s) Type(s) of plant Used model(s)/Algorithms/ Accuracy %
Technique(s)

✓ Soybean ✓ YoloV5 ✓ 98.75

✓ InceptionV3 ✓ 97.00
[1]

✓ CNN ✓ 97.00
✓ Banana ✓ CNN ✓ 93.45
✓ Rice ✓ ResNet 50 ✓ 91.68
[3]

✓ ResNet101 ✓ 92.50
[4]

✓ DenseNet161 ✓ 95.74
✓ DenseNet169 ✓ 94.98
✓ Tomato ✓ PCA, Linear SVM ✓ 88.67
✓ Potato ✓ LR ✓ 66.4
[5]

✓ Tomato ✓ KNN ✓ 54.5

[6]

✓ Strawberry ✓ CNN ✓ 53.4

✓ Corn ✓ SVM ✓ 98.0
✓ Grape
✓ Apple
✓ Apple ✓ CNN ✓ 96.5
✓ Corn
[7]

✓ Grapes
✓ Potato
✓ Sugarcane
✓ Tomato
✓ Tomato ✓ CNN and LQV ✓ 86.00
✓ Rice ✓ InceptionResNetV2 ✓ 98.9
[8]

✓ Xception ✓ 97.65
[9]

✓ ResNet50 ✓ 97.00
✓ MobileNet ✓ 96.65
✓ InceptionV3 ✓ 95.85
✓ Turmeric ✓ VGG-16 ✓ 96.24
✓ Different plant leaf species ✓ SVM ✓ 92.4
[12]

✓ Tomato ✓ CNN ✓ 98.029 (for testing)

[13]

✓ Pepper ✓ 98.29 (for training)

[14]

✓ Potato
✓ Rice ✓ DCNN ✓ 99.7
✓ Tomato ✓ CNN ✓ 99.64
[25]

✓ Rice ✓ DenseNet-121 ✓ 98.00

[26]

✓ Apple ✓ 96.00
[29]

✓ Bean ✓ 94.00
✓ Potato ✓ 95.00
✓ Tomato ✓ 97.00
✓ Tomato ✓ AlexNet with TL ✓ 89.69 with an 80/20 and 88.45%
✓ AlexNet with FE
[33]

✓ 93.4 with an 80/20 and 92.11%

with a 70/30 ratios

✓ Rice ✓ TL ✓ 80.00
with a 70/30 ratios

✓ CNN + TL ✓ 85.00
[39]

✓ ANN ✓ 90.00
✓ ECNN + GA ✓ 95.00
✓ Different plant leaf species ✓ LR ✓ 71.89
✓ SVM ✓ 75.76
[40]

✓ KNN ✓ 82.17
✓ RF ✓ 97.12
✓ NB ✓ 81.12
✓ CNN ✓ 98.43
✓ Rice ✓ DenseNet169 ✓ 99.66
✓ Xception (fine-tuned TL) ✓ 99.99
[41]

✓ Different plant leaf species ✓ C-GAN ✓ 99.6

✓ CNN ✓ 99.97
[42]

✓ SGD ✓ 85.00
✓ ACO-CNN ✓ 99.98
Demilie Journal of Big Data (2024) 11:5 Page 27 of 43

Table 1 (continued)
Author(s) Type(s) of plant Used model(s)/Algorithms/ Accuracy %
Technique(s)

✓ Different plant leaf species ✓ EfficientB5Net ✓ 94.512

✓ InceptionV3Net ✓ 96.347
[43]

✓ DenseNet201 ✓ 95.481
✓ AlexNet ✓ 89.548
✓ ResNet152 ✓ 95.728
✓ VGG19Net ✓ 92.695
✓ PPLC-Net ✓ 99.702
✓ Tomato ✓ 2-DCNN ✓ 96.00
✓ Apple ✓ Dilated TL and EL ✓ 99.10
[44]

✓ Corn
[45]

✓ Cotton
✓ Grape
✓ Pepper
✓ Rice
✓ Lemon ✓ SVM with the proposed ✓ 95.71
✓ Banana
[46]

✓ Beans
algorithm

✓ Rose
✓ Tomato ✓ EL based DL ✓ 96.00
✓ Tomato ✓ ResNet50-CBAM + SVM ✓ 97.20
[47]

✓ Tomato ✓ Faster-RCNN (RESNET-34) ✓ 99.97

[48]

✓ Ginger ✓ CNN ✓ 95.2

[49]

✓ Different plant leaf species ✓ DCNN with YOLOv7 ✓ 99.50

[51]

✓ Cotton ✓ CNN ✓ 100 and 90 for identification and

[52]
[53]

✓ Coffee ✓ DT with BPNN ✓ 94.5

classification respectively

✓ Tomato ✓ GAR​ ✓ 96.70

[54]

✓ Tomato ✓ Customized U-Net ✓ 98.12

[55]

✓ Coffee ✓ TL through Mobilnet ✓ 97.01

[56]

✓ Resnet50 ✓ 99.89
[57]

✓ Tomato ✓ U-Net and Modified U-Net ✓ 99.97 (binary class)

✓ 99.22 (Multi-Class (6))
[58]

✓ 99.91 (Multi-Class (10))

✓ Tomato ✓ KNN ✓ 99.92
✓ SVM ✓ 99.90
[59]

✓ Olive ✓ MobiRes-Net ✓ 97.08

✓ ResNet50 ✓ 94.86
[60]

✓ MobileNet ✓ 95.63
✓ Rice ✓ PlantDet ✓ 98.53
✓ Different plant leaf species ✓ DenseNet-77 ✓ 99.98
[61]

✓ Tomato ✓ DCNN ✓ 98.49

[62]

✓ Tomato ✓ A multinomial LR ✓ 97.00

[63]

✓ Canola ✓ K-FLBPCM ✓ 98.63

[64]

✓ Corn
[65]

✓ Wild radish
✓ Tea ✓ GLCM with Harris ✓ 97.48
✓ and SVM
[66]

✓ Turmeric ✓ KMC, GLCM, GLCM, and SVM ✓ 91.61

✓ Tomato ✓ CNN ✓ 94.00
[67]

✓ Maize ✓ SVM ✓ 77.56

[68]

✓ NB ✓ 77.46
[69]

✓ KNN ✓ 76.16
✓ DT ✓ 74.35
✓ RF ✓ 79.23
[70] ✓ Rice ✓ TL (InceptionV3, MobileNetV2 ✓ 96.42
and DenseNet121)
Demilie J ournal of Big Data (2024) 11:5 Page 28 of 43

Table 1 (continued)
Author(s) Type(s) of plant Used model(s)/Algorithms/ Accuracy %
Technique(s)

[71] ✓ Tomato ✓ ResNet101,VGG16,VGG19,Goo ✓ 94.6

✓ Tomato ✓ MobileNetV2 ✓ 99.30

gleLeNet, AlexNet, ResNet50

✓ Apple ✓ InceptionV3 ✓ 98.42

[72]

✓ Blueberry ✓ InceptionResnetV2 ✓ 99.11

[73]

✓ Cherry ✓ MobileNetV2 ✓ 97.02

✓ Corn ✓ EfficientNetB0 ✓ 99.56
✓ Grape
✓ Orange
✓ Peach
✓ Pepper bell
✓ Potato
✓ Raspberry
✓ Soybean
✓ Squash
✓ Strawberry
✓ Tomato
✓ Tomato ✓ SE-ResNet50 ✓ 96.81
✓ Tomato ✓ OpenCV ✓ 98.00
[74]

✓ Tomato ✓ DNN ✓ 86.18

[75]

✓ Tomato ✓ GoogleNet, AlexNet, Inception ✓ 99.72

[76]
[77]

✓ Tomato ✓ AlexNet ✓ 76.1

V3,ResNet 18,ResNet 50

✓ Tomato ✓ MobileNet V2 ✓ 90.00

[78]

✓ Tomato ✓ Hybrid SVM ✓ 92.37

[79]

✓ Apple ✓ DCNN ✓ 96.46

[80]

✓ Blueberry
[81]

✓ Cherry
✓ Corn
✓ Grape
✓ Orange
✓ Peach
✓ Pepper
✓ Potato
✓ Raspberry
✓ Soybean
✓ Squash
✓ Strawberry
✓ Tomato
✓ Tomato ✓ Multiclass SVM ✓ 94.00
✓ Tomato ✓ TL based DCNN ✓ 99.55
[82]

✓ Tomato ✓ VGG16 ✓ 91.2

[83]

✓ InceptionV3 ✓ 63.40
[84]

✓ MobileNet ✓ 63.75
✓ Tomato ✓ CNN in PlantVillage dataset ✓ 98.4
✓ CNN other than the PlantVil- ✓ 98.7
[85]

✓ 94.9
✓ A traditional ML with KNN ✓ 93.5
lage dataset

✓ VGG16
✓ Citrus ✓ GLCM, k-means, SVM ✓ 90.00
✓ Rice ✓ Vgg16 ✓ 70.42
[87]

✓ Vgg19 ✓ 73.60
[90]

✓ ResNet50 ✓ 51.99
✓ ResNet50v2 ✓ 61.60
✓ ResNet101v2 ✓ 86.79
✓ Rice ✓ SVM, DCNN ✓ 97.5
✓ Grape ✓ KMC and SVM ✓ 88.89
[91]

✓ Soybean ✓ SIFT and SVM ✓ 93.79

[93]
[94]
Demilie Journal of Big Data (2024) 11:5 Page 29 of 43

Table 1 (continued)
Author(s) Type(s) of plant Used model(s)/Algorithms/ Accuracy %
Technique(s)

✓ Maize ✓ MobileNet50 ✓ 74.90

✓ Grape ✓ PDDNN ✓ 86.00
[95]

✓ Apple
✓ Tomato
✓ Cotton ✓ DCNN ✓ 96.00
✓ Tomato ✓ AlexNet ✓ 95.65
[96]

✓ SqueezeNet ✓ 94.30
[98]

✓ Tomato ✓ CNN ✓ 97.00

✓ Rice ✓ CNN ✓ 92.46
[99]

✓ Guava ✓ DCNN ✓ 98.74

[105]

✓ Apple ✓ CNN ✓ 99.35

[110]

✓ Blueberry
[112]

✓ Cherry
✓ Corn
✓ Grape
✓ Orange
✓ Peach
✓ Pepper bell
✓ Potato
✓ Raspberry
✓ Soybean
✓ Squash
✓ Strawberry
✓ Tomato
✓ Cotton ✓ ANN ✓ 80.00
✓ Different plant leaf species ✓ ANN ✓ 92.00
[113]

✓ Groundnut ✓ Back propagation ✓ 97.00

[114]

✓ Phyllanthus Elegans ✓ MLP ✓ 90.15

[119]

✓ RBF ✓ 98.85
[121]

✓ Cotton ✓ ANN ✓ 83.00

✓ Soybeans ✓ 80.00
[122]

✓ Potato ✓ CNN, Bayesian algorithm ✓ 98.90

✓ Tomato
[123]

✓ Pepper bell
✓ Grapes ✓ Hybrid CNN ✓ 98.70
✓ Strawberry ✓ FL ✓ 96.00
[124]

✓ Apple ✓ PR ✓ 90.00
[125]

✓ Maize ✓ CNN ✓ 96.76

[126]

✓ Different plant leaf species ✓ NB ✓ 97.00

[128]

✓ Rice ✓ CNN ✓ 99.58

[132]

✓ Potato ✓ 97.66
[134]

✓ Guava ✓ CNN ✓ 95.61

✓ Potato ✓ KMC and GLCM ✓ 95.99
[136]

✓ Beans ✓ MobileNet ✓ 92.00

[137]

✓ Maize ✓ GoogleNet ✓ 98.90

[138]

✓ Cifa10 ✓ 98.80
[139]

✓ Onion ✓ DCNN ✓ 85.47

✓ Rice ✓ ADSNN-BO ✓ 94.65
[140]

✓ Tomato ✓ RF ✓ 95.00
[142]

✓ Different plant leaf species ✓ KMC and CNN ✓ 92.60

[143]

✓ Sugarcane ✓ SVM ✓ 95.00

[144]

✓ Cassava ✓ ANN and KNN ✓ 90.00

[145]
[146]
Demilie J ournal of Big Data (2024) 11:5 Page 30 of 43

Table 1 (continued)
Author(s) Type(s) of plant Used model(s)/Algorithms/ Accuracy %
Technique(s)

✓ Tomato ✓ ML and DL ✓ 99.4 for binary class

✓ Potato ✓ 99.2 for multiclass
[147]

✓ Rice
✓ Pepper bell
✓ Maize ✓ YOLOv3-tiny ✓ 69.40
✓ YOLOv4 ✓ 97.50
[148]

✓ YOLOv5s ✓ 88.23
✓ YOLOv7s ✓ 93.30
✓ YOLOv8n ✓ 99.04
✓ Rice ✓ ADLWNN ✓ 98.17
✓ Cofee ✓ KDE + ResNet50 ✓ 98.00
[149]

✓ Apple ✓ Res-ATTEN ✓ 99.00

[150]

✓ Rice ✓ 99.00
[151]

✓ Corn ✓ 94.00
✓ Grape ✓ 97.00
✓ Maiz ✓ DeepPlantNe DL ✓ 98.49 (in eight classes) and 99.85
✓ Potato
[152]

✓ Tomato
(in three classes)

✓ Apple ✓ PDD-Net ✓ 93.79 (in PlantVillege dataset)

✓ Maize ✓ 86.98 (only for Casava)
[153]

✓ Cherry
✓ Corn/Maze
✓ Grape
✓ Peach
✓ Potato
✓ Cassava
✓ Tomato ✓ CNN ✓ 99.60
✓ Sugarcane ✓ DNSVM ✓ 97.78
[154]

✓ Palm ✓ ResNet ✓ 99.62 (for the original dataset)

[155]
[156]
and 100% (for the augmented

✓ Maize ✓ RF ✓ 80.68
dataset)

✓ Mini-leaves ✓ SSM-Net ✓ 92.7

[157]

✓ Sugarcane ✓ 94.30
[158]

✓ Vegetables ✓ KMC ✓ 95.16

✓ Weed ✓ Histogram analysis
[159]

✓ SIFT ✓ 99.00
[160] 95.00

[161] ✓ Rice ✓ TL ✓ 99.64

techniques by using different performance evaluation metrics such as accuracy, mean

average precision (mAP), recall, precision, Intersection over Union (IoU), sensitiv-
ity, and specificity [62]. Equations (1), (2), (3), (4), (5), (6), (7), and (8) are defined and
describe how to evaluate the performances of the plant disease detection and classifica-
tion techniques.
A. Accuracy The accuracy score of a model, often known as accuracy, is a classification
statistic in DL and ML techniques that represents the proportion of correct predictions
made by the model.

TP + TN
Accuracy = (1)
TP + FP + TN + FN
Demilie Journal of Big Data (2024) 11:5 Page 31 of 43

25

20

15

10

0
2010 2013 2015 2016 2017 2018 2019 2020 2021 2022 2023
Distribution of plant leaf and/or crop disease detection and classification research papers per
year
Fig. 5 Distribution of plant leaf and/or crop disease detection and classification research papers per year (i.e.,
for this particular study)

B. mAP When the Intersection over Union (IOU) is greater than or equal to the
threshold, mAP is the accuracy. It is the ratio of a true and predicted region of inter-
ests’ intersection area to the union of the same. At different thresholds, detection perfor-
mance is measured in thresholds of mAP.
T AP(Ti)
mAP = (2)
i=1 T

where AP denotes the average precision of each class and q is the query or test
image. Q is the total number of test images:
C. Recall A model’s recall is defined as the model’s ability to correctly identify True
Positives. It is defined as the ratio of correctly classified positive outputs to correctly
classified outputs.

TP
Recall = (3)
(TP + FN )

D Precision Precision is defined as the ability to identify only relevant objects. It is

defined as the ratio of correctly classified positive outputs to total positive outputs.

TP
Precision = (4)
(TP + FP)

E. F1-score The f1 score is also introduced to assess the model’s accuracy. The f1-score
considers both the model’s precision and recall [34].

2(∗Precision ∗ Recall)
F 1 − score = 2 ∗ (Precision ∗ ∗ 100% (5)
(Precision + Recall)

F. intersection over union (IoU)

TP
IoU = ∗2 (6)
(FN + FP + TP)
Demilie J ournal of Big Data (2024) 11:5 Page 32 of 43

G. Sensitivity

TP
Sensitivity = (7)
(TP + FN )

H. Specificity

TN
Specificity = (8)
(TN + FP)

where TP, TN, FP, and FN are for True Positive, True Negative, False Positive, and False
Negative respectively.
In addition to the above performance evaluation metrics, determining the perfor-
mances of the plant leaf and /or crop disease detection and classification performance
should be evaluated in terms of the Logarithmic Loss (including the training loss, valida-
tion loss, and testing loss) and Area Under Curve (AUC) metrics accordingly. As indi-
cated in [62], designing an effective logarithmic loss function is mandatory for the robust
performance of plant leaf and /or crop disease detection and classification models.

Results and discussions

A comparative review of different research works done in different plant leaf disease
detection and classification using DL and ML techniques has been investigated by many
researchers. Accordingly, when sufficient data is available for training, DL techniques are
capable of detecting and classifying plant leaf diseases with high accuracy. As indicated
in Figs. 6, 7, 8, and 9, different researchers have developed plant leave disease detec-
tion and classification systems by using plant leave images. Here, most of the research-
ers have tested their proposed systems on different plant leaves. They have computed
the evaluation metrics such as accuracy, precision, recall, f1-score, mAP, IoU, sensitiv-
ity, specificity, and Matthews correlation coefficient (MCC) for training and testing pur-
poses. Accordingly, for this particular comparative study, the author has considered only

Accuracy
99.99%

99.98%
99.97%
99.70%

99.66%
99.64%

99.60%
98.90%
98.75%

98.43%
98.29%
98.00%

97.65%

97.12%
96.65%
96.50%

96.24%
95.85%
95.74%
94.98%
93.45%

93.40%
92.50%

92.40%
91.68%

98%
97%
97%

97%

89.69%
88.67%

95%
90%

82.17%

81.12%
86%

85%

85%
75.76%
80%

71.89%
66.40%
54.50%
53.40%

Fig. 6 Performances of the DL and ML techniques in the study a

Demilie Journal of Big Data (2024) 11:5 Page 33 of 43

Accuracy

99.98%
99.97%

99.92%
99.91%

99.90%
99.89%
99.70%

99.50%
99.10%

98.63%
98.53%

98.49%
98.12%

97.48%
97.20%

97.08%
97.01%
96.70%

96.42%
96.35%

100%
95.73%

95.71%

95.63%
95.48%

95.20%

94.86%

94.60%
94.51%

94.50%
92.70%

91.61%
97%
96%

96%
89.55%

94%

79.23%
77.56%
77.46%
76.16%
74.35%
Fig. 7 Performances of the DL and ML techniques in the study b

Accuracy
99.72%
99.56%

99.55%

99.35%
99.30%

99.11%

98.74%
98.70%
98.42%

98.40%

97.50%
97.02%

96.81%

96.46%

95.65%
94.90%

94.30%
93.79%
93.50%

92.46%
92.37%

91.20%

88.89%
98%

97%
86.79%

96%
86.18%

94%
90%

90%

86%
76.10%

74.90%
73.60%
70.42%

80%
63.75%
63.40%

61.60%
51.99%

Fig. 8 Performances of the DL and ML techniques in the study c

the accuracy of the literature. Additionally, the maximum accuracy (including multiclass
level accuracies accordingly) result has been considered in plant leaf disease detection
and classification research work that has been tested on different plant leaves in the
literature.
As depicted in Figs. 6, 7, 8, and 9, CNNs are generally superior to SVMs for plant
leaf disease detection and classification tasks by using the images of plants accordingly.
Demilie J ournal of Big Data (2024) 11:5 Page 34 of 43

Accuracy

99.85%

99.64%
99.60%
99.58%

99.20%

99.04%

99.00%
98.90%
98.90%
98.85%

98.80%
98.70%

98.17%
98.00%

97.78%
97.50%
96.76%

95.99%
95.61%

95.16%
100%
94.65%

94.30%
93.79%
93.30%
92.60%

99%
90.15%

88.23%
97%
97%

96%

95%

95%

95%
85.47%
92%

92%
90%

90%

80.68%
83%

69.40%
Fig. 9 Performances of the DL and ML techniques in the study d

CNNs are specifically designed for handling image data, as they can automatically
learn hierarchical features from raw pixel values. They use convolutional layers to cap-
ture local patterns and hierarchical representations, enabling them to excel at tasks like
object recognition, image detection, and classification. DL and ML both have roles in
image detection and classification. ML techniques like SVMs may be sufficient for sim-
ple image detection and classification tasks with well-defined features. However, DL,
including CNNs, is typically more powerful and versatile for complex image tasks. DL
models can automatically learn intricate features and representations, reducing the need
for manual feature engineering. Hence, for most modern image detection classification
tasks, DL, particularly CNNs, is the preferred choice due to their ability to outperform
traditional ML techniques in terms of accuracy and efficiency [161].
CNNs are often the favored choice for image detection and classification due to their
inherent capacity to autonomously acquire pertinent image features and grasp spatial
hierarchies. Nevertheless, the selection between conventional ML and DL hinges on the
particular problem, the accessibility of data, and the computational capabilities accessi-
ble. In numerous advanced image detection and classification tasks, DL, mainly through
CNNs, is preferred when ample data and computational resources are available.
The significance of collecting large datasets with high variability, data augmentation,
transfer learning, and visualization of CNN activation maps in improving detection and
classification accuracies, as well as the significance of small sample plant leaf disease
detection, classification, and hyper-spectral imaging for early detection and classifica-
tion of plant disease, have all been mandatory activities to increase the best results of
agricultural products [62]. Most of the DL frameworks proposed in the study showed
good detection and classification effects on their datasets but not on other datasets, indi-
cating that the model is not good. As a result, more robust DL models are required to
adapt to the different disease datasets. The PlantVillage dataset was used to evaluate the
performance of both DL and ML techniques in most of the studies. Although the data-
set contains several images of various plant species with diseases, they were all taken in
the lab. As a result, a large dataset of plant diseases in real-world situations is expected.
Although some researchers use hyperspectral images of diseased leaves and different
Demilie Journal of Big Data (2024) 11:5 Page 35 of 43

DL and ML frameworks for the early detection and classification of plant leaf diseases,
issues that affect the widespread use of hyperspectral imaging (HSI) in the early detec-
tion and classification of plant diseases have yet to be resolved [168]. That is, it is difficult
to get labeled datasets for early plant disease detection and classification, and even expe-
rienced experts cannot mark where the invisible disease symptoms are and define purely
invisible disease pixels, which is critical for HSI to detect plant disease. The analysis of
the reviewed papers revealed that the detection and classification of plant disease is a
hot issue and needs further investigation. The detection and classification of plant leaf
disease for different crops are a pressing issue and a complex process, partly due to the
availability of datasets and the task of developing an ML model with rigid performance.
As depicted in Figs. 6, 7, 8, and 9 below, based on the reviewed research works in the
area, most researchers have achieved better evaluation results with the DL technique
than with the ML technique for developing plant disease detection and classification
systems. Additionally, most researchers have combined different techniques, such as DL
and ML, to detect and classify various plant leaves. Accordingly, though SVM has excel-
lent performance in image detection and classification, its performance depends on the
image features it receives. It is suggested to combine CNN features with the SVM classi-
fier unless the researchers plan to use pre-trained models.
As depicted in Fig. 10, most of the studies that have been included in this particu-
lar study have concentrated on the detection and classification of different diseases
in tomato and rice crops, respectively. In addition to the tomato and rice crops, the
researchers have tried to conduct disease detection and classification studies on differ-
ent plant leaves and/or crop species accordingly.

Soybean
1% 1% 2%
1% 1% 1% Banana
2%
1% Rice
1% Tomato
1%
1% 16% Turmeric
1%
1% Ginger
1% Cotton
2% Coffee
Olive
2% Tea
1% Maize
Citrus
Grape
6%
Guava
Groundnut
Strawberry
1%
1% Apple
Phyllanthus elegans
4% Potato
Beans
4% Onion
Sugarcane
1%
39% Cassava
2%
Vegetables
Weed
Palm

Fig. 10 Distribution of the studies per plant leaf and/or crop species (i.e., for this particular study)
Demilie J ournal of Big Data (2024) 11:5 Page 36 of 43

Apple, Maize, Cherry,

Corn/Maze, Grape, Peach,
Maiz, Potato, and Tomato Potato, and Cassava
1% 1%
Potato, Tomato, Strawberry,
Apple, Rice, Corn , and Grape Corn, Grape, and Apple
1%
Mini-leaves and Sugarcane 9%
3%
Rice and Potato
3%

Potato, Tomato, and Pepper bell Apple, Corn, Grapes, Potato,

4% Sugarcane, and Tomato
9%

Cotton and Soybeans

3%

Maize, Grape, Apple, Tomato, Pepper,

and Tomato and Potato
6% 4%

Rice, Apple, Bean,

Tomato, Potato, Rice, Potato, and Tomato
and Pepper bell 7%
6%

Apple, Corn, Cotton,

Grape, Pepper, and Rice
9%

Apple, Blueberry, Cherry, Corn,

Grape, Orange, Peach, Pepper Lemon, Banana, Beans, and
bell, Potato, Raspberry, Rose
Soybean, Squash, Strawberry, 6%
and Tomato
Canola, Corn, and Wild radish
21%
4%

Fig. 11 Distribution of the studies per multiple plant leaf and/or crop species

Currently, plant leaf and/or crop species disease detection and classification studies
have been implemented and tested using different techniques and approaches for multi-
ple plant leaf and/or crop species accordingly. As summarized in Fig. 11, plant leaf and/
or crop species disease detection and classification techniques and approaches have
been implemented for more than two plants. From Fig. 11, the Apple, Blueberry, Cherry,
Corn, Grape, Orange, Peach, Pepper Bell, Potato, Raspberry, Soybean, Squash, Straw-
berry, and Tomato plant leaf and/or crop species have been implemented and tested by
the CNN technique and achieved an overall accuracy of 99.35%.
Additionally, most of the researchers have tried to include the tomato plant in the
selected plant leaf and/or crop species disease detection and classification research
works. This indicates that the tomato plant is the most researchable plant leaf and/or
crop species.

Conclusion and recommendation

The literature provided many strategies that have been developed and made available to
aid in the achievement of enormous accomplishments in the fields of ML, DL, and image
processing. The assessment also stated that the percentage accuracy can be enhanced by
training and testing the models with more and more datasets to boost the classification
and identification accuracy. To address the problem of plant leaf or crop disease, new
Demilie Journal of Big Data (2024) 11:5 Page 37 of 43

and better DL algorithms that can provide higher accuracy in identifying and classifying
plant leaf or crop diseases must always be developed and applied.
Several methods for classifying plant leaf diseases have been developed and used in
the past. However, neural networks, such as CNN, appear to be the ideal technique
for classifying plant diseases due to their flexibility and feature extractor property,
which allows them to extract features automatically. Unlike prior models such as NB,
KNN, SVM, RFC, and others, CNN may learn extra features from images to produce
superior results. Because of their sophistication in learning and extracting infor-
mation from images for trustworthy output, neural networks such as CNN are well
suited for research work in the areas of computer vision and image processing. In
addition, from the DL and ML techniques that have been included in this particular
study, CNNs are often the favored choice for image detection and classification due to
their inherent capacity to autonomously acquire pertinent image features and grasp
spatial hierarchies. Nevertheless, the selection between conventional ML and DL
hinges upon the particular problem, the accessibility of data, and the computational
capabilities accessible. Accordingly, in numerous advanced image detection and clas-
sification tasks, DL, mainly through CNNs, is preferred when ample data and compu-
tational resources are available and show good detection and classification effects on
their datasets, but not on other datasets.
Finally, the author recommends future researchers to investigate plant disease
detection and classification research works by considering the following points
accordingly:

• Developing DL models for disease detection and classification in different parts of

plants.
• To improve the robustness of homemade datasets in complicated contexts,
an automated parameter search technique for the weather data augmentation
method should be developed.
• Data augmentation, large datasets with significant variability, and other strategies
can increase plant disease detection and classification accuracies.
• Expanding the recommended techniques will result in a significant contribution
to sustainable agriculture, influencing crop quality for future generations.
• Using preprocessing techniques on the dataset, such as resizing and augmenta-
tion.
• DL models for real-time disease detection are being developed.
• Development of an Android app that detects the presence of diseases on differ-
ent plants using a DL model.
• Develop such plant leave detection and classification systems by combining DL and
ML techniques such as DCCN and CNN features with SVM techniques.

Acknowledgements
I want to thank all the researchers who have been cited (included) in this particular research work.

Author contributions
WBD: Prepared manuscript including analysis, data curation, visualization, conceptualization, methodology, and writing
of the original draft.
Demilie J ournal of Big Data (2024) 11:5 Page 38 of 43

Funding
This study did not receive any funds.

Availability of data and materials

The research data are available from the authors and can be accessed from the author accordingly.

Declarations
Ethics approval and consent to participate
Not applicable.

Consent for publication

Not applicable.

Competing interests
The author declares that no competing interest.

Received: 10 September 2023 Accepted: 14 December 2023

References
1. Tirkey D, Singh KK, Tripathi S. Performance analysis of AI-based solutions for crop disease identification detection,
and classification. Smart Agric Technol. 2023. https://​doi.​org/​10.​1016/j.​atech.​2023.​100238.
2. Ramanjot, et al. Plant disease detection and classification: a systematic literature review”. Sensors. 2023. https://​doi.​
org/​10.​3390/​s2310​4769.
3. Krishnan VG, Deepa J, Rao PV, Divya V, Kaviarasan S. An automated segmentation and classification model for
banana leaf disease detection. J Appl Biol Biotechnol. 2022;10(1):213–20. https://​doi.​org/​10.​7324/​JABB.​2021.​
100126.
4. S Mathulaprangsan K Lanthong S Patarapuwadol. 2020. Rice Diseases Recognition Using Effective Deep Learning
Models. Telecommun. Eng Media Technol with ECTI North Sect Conf Electr Electron Jt Int Conf Digit Arts Comput.
https://​doi.​org/​10.​1109/​ECTID​AMTNC​ON482​61.​2020.​90907​09
5. Heltin Genitha C, Dhinesh E, Jagan A. Detection of leaf disease using principal component analysis and linear sup-
port vector machine. Adv Comput ICoAC Proc Int Conf. 2019. https://​doi.​org/​10.​1109/​ICoAC​48765.​2019.​246866.
6. P. S. Gupta, P. Hans, and S. Chand. 2020. Classification Of Plant Leaf Diseases Using Machine Learning And Image
Preprocessing Techniques.
7. S. V. Militante, B. D. Gerardo, and N. V. Dionisio, “Plant Leaf Detection and Disease Recognition using Deep Learn-
ing,” 2019 IEEE Eurasia Conf. IOT, Commun. Eng., pp. 579–582, 2019.
8. Sardogan M, Tuncer A, Ozen Y. Plant leaf disease detection and classification based on CNN with LVQ algorithm.
Comput Sci Eng Conf. 2018. https://​doi.​org/​10.​1109/​UBMK.​2018.​85666​35.
9. Sharma M, Kumar CJ, Deka A. Early diagnosis of rice plant disease using machine learning techniques. Arch Phyto-
pathol Plant Prot. 2022;55(3):259–83. https://​doi.​org/​10.​1080/​03235​408.​2021.​20158​66.
10. Sharath DM, Akhilesh S, Kumar A, Rohan MG, Prathap C. Image based plant disease detection in pomegranate
plant for bacterial blight. Proc IEEE Int Conf Commun Signal Process ICCSP. 2019. https://​doi.​org/​10.​1109/​ICCSP.​
2019.​86980​07.
11. Kumar S, Raghavendra BK. Diseases detection of various plant leaf using image processing techniques: a review”
2019 ­5th. Comput Commun Int Conf Adv. 2019. https://​doi.​org/​10.​1109/​ICACCS.​2019.​87283​25.
12. Rajasekaran C, Arul S, Devi S, Gowtham G, Jeyaram S. Turmeric plant diseases detection and classification using
artificial intelligence. Int Conf Signal Process Commun. 2020. https://​doi.​org/​10.​1109/​ICCSP​48568.​2020.​91822​55.
13. P. V., R. Das, and K. V. Identification of plant leaf diseases using image processing techniques. 2017 IEEE Int. Conf.
Technol. Innov. ICT Agric. Rural Dev. (TIAR 2017), pp. 130–133, 2017.
14. Jasim MA, Al-Tuwaijari JM. Plant leaf diseases detection and classification using image processing and deep learn-
ing techniques. Int Comput Sci Soft Eng Conf. 2020. https://​doi.​org/​10.​1109/​CSASE​48920.​2020.​91420​97.
15. Bedi P, Gole P. Plant disease detection using hybrid model based on convolutional autoencoder and convolutional
neural network. Artif Intell Agric. 2021;5:90–101. https://​doi.​org/​10.​1016/j.​aiia.​2021.​05.​002.
16. Uchida S, Ide S, Iwana BK, Zhu A. A further step to perfect accuracy by training CNN with larger data. Int Conf
Front Handwrit Recognit. 2016. https://​doi.​org/​10.​1109/​ICFHR.​2016.​0082.
17. Hu Y, Liu G, Chen Z, Liu J, Guo J. Lightweight one-stage maize leaf disease detection model with knowledge distil-
lation. Agriculture. 2023;13:1–22.
18. Ma L, Yu Q, Yu H, Zhang J. Maize leaf disease identification based on yolov5n algorithm incorporating attention
mechanism. Agronomy. 2023. https://​doi.​org/​10.​3390/​agron​omy13​020521.
19. Kumar R, Chug A, Singh AP, Singh D. A systematic analysis of machine learning and deep learning based
approaches for plant leaf disease classification: a Review. J Sensors. 2022. https://​doi.​org/​10.​1155/​2022/​32875​61.
20. Saleem MH, Potgieter J, Arif KM. Plant disease classification: a comparative evaluation of convolutional neural
networks and deep learning optimizers. Plants. 2020;9(10):1–17. https://​doi.​org/​10.​3390/​plant​s9101​319.
21. Tiwari V, Joshi RC, Dutta MK. Dense convolutional neural networks based multiclass plant disease detection and
classification using leaf images. Ecol Inform. 2021;63: 101289. https://​doi.​org/​10.​1016/j.​ecoinf.​2021.​101289.
22. Vishnoi VK, Kumar K, Kumar B. Plant disease detection using computational intelligence and image processing.
Berlin Heidelberg: Springer; 2021.
Demilie Journal of Big Data (2024) 11:5 Page 39 of 43

23. Faizal Azizi MM, Lau HY. Advanced diagnostic approaches developed for the global menace of rice diseases: a
review. Can J Plant Pathol. 2022;44(5):627–51. https://​doi.​org/​10.​1080/​07060​661.​2022.​20535​88.
24. Shoaib M, et al. An advanced deep learning models-based plant disease detection: a review of recent research.
Front Plant Sci. 2023;14:1–22. https://​doi.​org/​10.​3389/​fpls.​2023.​11589​33.
25. Upadhyay SK, Kumar A. A novel approach for rice plant diseases classification with deep convolutional neural
network. Int J Inf Technol. 2022;14(1):185–99. https://​doi.​org/​10.​1007/​s41870-​021-​00817-5.
26. Guerrero-Ibañez A, Reyes-Muñoz A. Monitoring tomato leaf disease through convolutional neural networks.
Electron. 2023;12(1):1–15. https://​doi.​org/​10.​3390/​elect​ronic​s1201​0229.
27. Picon A, Seitz M, Alvarez-Gila A, Mohnke P, Ortiz-Barredo A, Echazarra J. Crop conditional convolutional neural
networks for massive multi-crop plant disease classification over cell phone acquired images taken on real field
conditions. Comput Electron Agric. 2019;167:105093. https://​doi.​org/​10.​1016/j.​compag.​2019.​105093.
28. Barbedo JGA. A review on the main challenges in automatic plant disease identification based on visible range
images. Biosyst Eng. 2016;144:52–60. https://​doi.​org/​10.​1016/j.​biosy​stems​eng.​2016.​01.​017.
29. Ahmed I, Yadav PK. A systematic analysis of machine learning and deep learning based approaches for identifying
and diagnosing plant diseases. Sustain Oper Comput. 2023;4:96–104. https://​doi.​org/​10.​1016/j.​susoc.​2023.​03.​001.
30. Dhiman P, Kaur A, Balasaraswathi VR, Gulzar Y, Alwan AA, Hamid Y. Image acquisition, preprocessing and classifica-
tion of citrus fruit diseases: a systematic literature review. Sustainability. 2023;15(12):9643. https://​doi.​org/​10.​3390/​
su151​29643.
31. Ramesh S, Vydeki D. Application of machine learning in detection of blast disease in south indian rice crops. J
Phytol. 2019;11:31–7. https://​doi.​org/​10.​25081/​jp.​2019.​v11.​5476.
32. Kc K, Yin Z, Li D, Wu Z. “Impacts of background removal on convolutional neural networks for plant disease clas-
sification in-situ. Agric. 2021. https://​doi.​org/​10.​3390/​agric​ultur​e1109​0827.
33. Verma S, Chug A, Singh AP. Application of convolutional neural networks for evaluation of disease severity in
tomato plant. J Discret Math Sci Cryptogr. 2020;23(1):273–82. https://​doi.​org/​10.​1080/​09720​529.​2020.​17218​90.
34. Liu J, Wang X. Plant diseases and pests detection based on deep learning: a review. Plant Methods. 2021;17(1):1–
18. https://​doi.​org/​10.​1186/​s13007-​021-​00722-9.
35. Wspanialy P, Moussa M. “A detection and severity estimation system for generic diseases of tomato greenhouse
plants. Comput Electron Agric. 2020;178:105701. https://​doi.​org/​10.​1016/j.​compag.​2020.​105701.
36. Reddy T, Dharmendra G, Rajput S, Kumar MP, Kuruva R. A novel PCA—whale optimization—based deep neural
network model for classification of tomato plant diseases using GPU. J Real-Time Image Process. 2021;18(4):1383–
96. https://​doi.​org/​10.​1007/​s11554-​020-​00987-8.
37. Chen J, Chen J, Zhang D, Nanehkaran YA, Sun Y. A cognitive vision method for the detection of plant disease
images. Mach Vis Appl. 2021;32(1):1–18. https://​doi.​org/​10.​1007/​s00138-​020-​01150-w.
38. Hossain SMM, Deb K, Dhar PK, Koshiba T. Plant leaf disease recognition using depth-wise separable convolution-
based models. Symmetry. 2021;13(3):1–29. https://​doi.​org/​10.​3390/​sym13​030511.
39. Balaji V, et al. Deep transfer learning technique for multimodal disease classification in plant images. Contrast
Media Mol Imaging. 2023;2023:5644727. https://​doi.​org/​10.​1155/​2023/​56447​27.
40. Kirola M, Joshi K, Chaudhary S, Singh N, Anandaram H, Gupta A. Plants diseases prediction framework: a image-
based system using deep learning. Proc IEEE World Conf Appl Intell Comput. 2022. https://​doi.​org/​10.​1109/​AIC55​
036.​2022.​98488​99.
41. S. C. Gopi and H. K. Kondaveeti. Transfer learning for rice leaf disease detection. 2023. https://​doi.​org/​10.​1109/​ICAIS​
56108.​2023.​10073​711.
42. Algani YMA, Caro OJM, Bravo LMR, Kaur C, Al Ansari MS, Bala BK. Leaf disease identification and classification using
optimized deep learning. Meas Sensors. 2023;25:100643. https://​doi.​org/​10.​1016/j.​measen.​2022.​100643.
43. Dai G, Fan J, Tian Z, Wang C. PPLC-Net : neural network-based plant disease identification model supported
by weather data augmentation and multi-level attention mechanism. J King Saud Univ Comput Inf Sci.
2023;35(5):101555. https://​doi.​org/​10.​1016/j.​jksuci.​2023.​101555.
44. P. B. R and A. V V. 2022. Tomato leaf disease detection and classification using Cnn. 71 4 2921–2930
45. M. S. Anari. A hybrid model for leaf diseases classification based on the modified deep transfer learning and
ensemble approach for agricultural aiot-based monitoring,” vol. 2022.
46. Singh V, Misra AK. Detection of plant leaf diseases using image segmentation and soft computing techniques
Process. Inf Agric. 2016. https://​doi.​org/​10.​1016/j.​inpa.​2016.​10.​005.
47. E. Saraswathi and J. FarithaBanu. A novel ensemble classification model for plant disease detection based on leaf
images. 2023.
48. Altalak M, Uddin MA, Alajmi A, Rizg A. A hybrid approach for the detection and classification of tomato leaf dis-
eases. Appl Sci. 2022;12:1–14.
49. Nawaz M, et al. OPEN A robust deep learning approach for tomato plant leaf disease localization and classification.
Rep Sci. 2022. https://​doi.​org/​10.​1038/​s41598-​022-​21498-5.
50. Garg S, Singh P. An aggregated loss function based lightweight few shot model for plant leaf disease classification.
Multimedia Tools Appl. 2023;10(23797–23815):2023.
51. Yigezu MG, Woldeyohannis MM, Tonja AL. Early ginger disease detection using deep learning approach. Berlin:
Springer International Publishing; 2022.
52. P. Nayar, S. Chhibber, and A. K. Dubey, “An Efficient Algorithm for Plant Disease Detection Using Deep Convo-
lutional Networks,” Proceedings - 2022 14th IEEE International Conference on Computational Intelligence and
Communication Networks, CICN 2022. pp. 156–160, 2022
53. Kukadiya H, Meva D. Automatic cotton leaf disease classification and detection by convolutional neural network.
Berlin: Springer Nature Switzerland; 2022.
54. Mengistu AD, Mengistu SG, Melesew D. An automatic coffee plant diseases identification using hybrid approaches
of image processing and decision tree. Indones J Electr Eng Comput Sci. 2018;9(3):806–11. https://​doi.​org/​10.​
11591/​ijeecs.​v9.​i3.​pp806-​811.
Demilie J ournal of Big Data (2024) 11:5 Page 40 of 43

55. Darma Putra IKG, Jayantha Putra IPD, Fauzi R. Witarsyah classification of tomato plants diseases using convolu-
tional neural network. Int J Adv Sci Eng Inf Technol. 2020;10(5):1821–7.
56. Gangwar A, Rani G, Dhaka VPS. detecting tomato crop diseases with Ai: leaf segmentation and analysis. Int Conf
Trends Electron Inform. 2023. https://​doi.​org/​10.​1109/​ICOEI​56765.​2023.​10125​891.
57. E. B. Paulos and M. M. Woldeyohannis. Detection and classification of coffee leaf disease using deep learning,” no.
January 2023, 2022. https://​doi.​org/​10.​1109/​ICT4D​A56482.​2022.​99713​00.
58. Shoaib M, et al. Deep learning-based segmentation and classification of leaf images for detection of tomato plant
disease. Front Plant Sci. 2022;13:1–18. https://​doi.​org/​10.​3389/​fpls.​2022.​10317​48.
59. Attallah O. Tomato leaf disease classification via compact convolutional neural networks with transfer learning and
feature selection. Horticulturae. 2023. https://​doi.​org/​10.​3390/​horti​cultu​rae90​20149.
60. Ksibi A, Ayadi M, Soufiene BO, Jamjoom MM, Ullah Z. MobiRes-net: a hybrid deep learning model for detecting
and classifying olive leaf diseases. Appl Sci. 2022. https://​doi.​org/​10.​3390/​app12​20102​78.
61. Shovon MSH, Mozumder SJ, Pal OK, Mridha MF, Asai N, Shin J. PlantDet: a robust multi-model ensemble method
based on deep learning for plant disease detection. IEEE Access. 2023;11(April):34846–59. https://​doi.​org/​10.​1109/​
ACCESS.​2023.​32648​35.
62. Albattah W, Nawaz M, Javed A, Masood M, Albahli S. A novel deep learning method for detection and classifica-
tion of plant diseases. Complex Intell Syst. 2022;8(1):507–24. https://​doi.​org/​10.​1007/​s40747-​021-​00536-1.
63. Trivedi NK, et al. Early detection and classification of tomato leaf disease using high-performance deep neural
network. Sensors. 2021. https://​doi.​org/​10.​3390/​s2123​7987.
64. Al-gaashani MSAM, Shang F, Muthanna MSA, Khayyat M, El-Latif AAA. Tomato leaf disease classification by exploit-
ing transfer learning and feature concatenation”. IET Image Process. 2022. https://​doi.​org/​10.​1049/​ipr2.​12397.
65. Thanh LVN, Ahderom S, Apopei B, Alameh K. A novel method for detecting morphologically similar crops and
weeds based on the combination of contour masks and filtered Local Binary Pattern operators. Gigascience.
2020;9(3):1–16. https://​doi.​org/​10.​1093/​gigas​cience/​giaa0​17.
66. Ahmad W, Shah SMA, Irtaza A. Plants disease phenotyping using quinary patterns as texture descriptor. KSII Trans
Internet Inf Syst. 2020;14(8):3312–27. https://​doi.​org/​10.​3837/​tiis.​2020.​08.​009.
67. G. Kuricheti and P. Supriya. Computer vision based turmeric leaf disease detection and classification: a step to
smart agriculture. 2019. https://​doi.​org/​10.​1109/​ICOEI.​2019.​88627​06.
68. A. Abbas, S. Jain, M. Gour, and S. Vankudothu. Tomato plant disease detection using transfer learning with C-GAN
synthetic images. 2021.
69. Panigrahi KP, Das H, Sahoo AK. Maize leaf disease detection and classification using machine learning maize
leaf disease detection and classification using machine intell. Adv Syst Comput. 2020. https://​doi.​org/​10.​1007/​
978-​981-​15-​2414-1.
70. Kanuparthi P, Bejgam V, Viswanatham VM. A novel approach of ensembling the transfer learning methods for rice
plant disease detection and classification. WEB. 2021. https://​doi.​org/​10.​14704/​WEB/​V18I2/​WEB18​331.
71. Prabhakar M, Purushothaman R, Awasthi DP. Deep learning based assessment of disease severity for early blight in
tomato crop. Multimed Tools Appl. 2020;79(39–40):28773–84. https://​doi.​org/​10.​1007/​s11042-​020-​09461-w.
72. Ahmed S, Hasan MB, Ahmed T, Sony MRK, Kabir MH. Less is more: lighter and faster deep neural architecture for
tomato leaf disease classification. IEEE Access. 2022;10(June):68868–84. https://​doi.​org/​10.​1109/​ACCESS.​2022.​
31872​03.
73. Hassan SM, Maji AK, Jasiński M, Leonowicz Z, Jasińska E. “Identification of plant-leaf diseases using CNN and
transfer-learning approach. Electronics. 2021. https://​doi.​org/​10.​3390/​elect​ronic​s1012​1388.
74. Zhao S, Peng Y, Liu J, Wu S. Tomato leaf disease diagnosis based on improved convolution neural network by
attention module. Agric. 2021. https://​doi.​org/​10.​3390/​agric​ultur​e1107​0651.
75. Ashok S, Kishore G, Rajesh V, Suchitra S, Gino Sophia SG, Pavithra B. Tomato leaf disease detection using deep
learning techniques. Proc 5th Int Conf Commun Electron Syst. 2020. https://​doi.​org/​10.​1109/​ICCES​48766.​2020.​
09137​986.
76. Sreelatha P, Udayakumar MS, Karthick S, Ch SC, Kavya KCS, Madiajagan M. Managing the tomato leaf dis-
ease detection accuracy using computer vision based deep neural network. J Contemp Issues Bus Gov.
2021;27(1):3425–37.
77. Maeda-Gutiérrez V, et al. Comparison of convolutional neural network architectures for classification of tomato
plant diseases. Appl Sci. 2020. https://​doi.​org/​10.​3390/​app10​041245.
78. Batool A, Hyder SB, Rahim A, Waheed N, Asghar MA. Classification and identification of tomato leaf disease using
deep neural network. 2020 Int Conf Eng Emerg Technol ICEET. 2020. https://​doi.​org/​10.​1109/​ICEET​48479.​2020.​
90482​07.
79. Zaki SZM, Zulkifley MA, Mohd Stofa M, Kamari NAM, Mohamed NA. Classification of tomato leaf diseases using
mobilenet v2. IAES Int J Artif Intell. 2020. https://​doi.​org/​10.​11591/​ijai.​v9.​i2.​pp290-​296.
80. Bhatia A, Chug A, Singh AP. Hybrid SVM-LR classifier for powdery mildew disease prediction in tomato plant. Int
Conf Signal Process Integr Networks. 2020. https://​doi.​org/​10.​1109/​SPIN4​8934.​2020.​90712​02.
81. Geetharamani G. Identification of plant leaf diseases using a nine-layer deep convolutional neural network. Com-
put Electr Eng. 2019;76:323–38. https://​doi.​org/​10.​1016/j.​compe​leceng.​2019.​04.​011.
82. Aziz S, Bashir M, Mughal O, Khan MU, Khan A. Image pattem classification for plant disease identification using
local tri-directional featuresInf. Electron Mob Commun Conf IEMCON Technol. 2029. https://​doi.​org/​10.​1109/​
IEMCON.​2019.​89362​36.
83. Thangaraj R, Anandamurugan S, Kaliappan VK. Automated tomato leaf disease classification using transfer
learning-based deep convolution neural network. J Plant Dis Prot. 2021;128(1):73–86. https://​doi.​org/​10.​1007/​
s41348-​020-​00403-0.
84. Agarwal M, Singh A, Arjaria S, Sinha A, Gupta S. ToLeD: tomato leaf disease detection using convolution neural
network. Procedia Comput Sci. 2020;167(2019):293–301. https://​doi.​org/​10.​1016/j.​procs.​2020.​03.​225.
85. Agarwal M, Gupta SK, Biswas KK. Development of efficient CNN model for tomato crop disease identification.
Sustain Comput Informatics Syst. 2020;28: 100407. https://​doi.​org/​10.​1016/j.​suscom.​2020.​100407.
Demilie Journal of Big Data (2024) 11:5 Page 41 of 43

86. Sandhu GK, Kaur R. “Plant disease detection techniques: a review”, 2019. Int Conf Autom Comput Technol Manag
ICACTM. 2019;2019:34–8. https://​doi.​org/​10.​1109/​ICACTM.​2019.​87768​27.
87. Meena Prakash R, Saraswathy GP, Ramalakshmi G, Mangaleswari KH, Kaviya T. Detection of leaf diseases and
classification using digital image processing. Int Conf Innov Information Embed Commun Syst ICIIECS Proc. 2018.
https://​doi.​org/​10.​1109/​ICIIE​CS.​2017.​82759​15.
88. M. Rashid, B. Ram, R. S. Batth, N. Ahmad, and H. M. E. I. Dafallaa. Novel image processing technique for feature
detection of wheat crops using python openCV.
89. Krithika P, Veni S. “Leaf disease detection on cucumber leaves using multiclass support vector machine. Proc Int
Conf Wirel Commun Signal Process Networking WiSPNET. 2017. https://​doi.​org/​10.​1109/​WiSPN​ET.​2017.​82999​69.
90. Burhan SA, Minhas DS, Tariq DA, Nabeel Hassan M. Comparative study of deep learning algorithms for disease and
pest detection in rice crops. Proc Conf Electron Comput Artif Intell ECAI Int. 2020. https://​doi.​org/​10.​1109/​ECAI5​
0035.​2020.​92232​39.
91. Hasan MJ, Mahbub S, Alom MS, Abu Nasim M. Rice disease Identification and classification by integrating support
vector machine with deep convolutional neural network. 1st Int Conf Adv Sci Eng Robot Technol. 2019. https://​
doi.​org/​10.​1109/​ICASE​RT.​2019.​89345​68.
92. Vaishnnave MP, Suganya Devi K, Srinivasan P, Arutperumjothi G. Detection and classification of groundnut leaf
diseases using KNN classifier. Int Conf Syst Comput Autom Networking ICSCAN. 2019. https://​doi.​org/​10.​1109/​
ICSCAN.​2019.​88787​33.
93. Padol PB, Yadav AA. SVM classifier based grape leaf disease detection. Conf Adv Signal Process CASP.
2016;2016:175–9. https://​doi.​org/​10.​1109/​CASP.​2016.​77461​60.
94. Dandawate Y, Kokare R. An automated approach for classification of plant diseases towards development of futur-
istic decision support system in Indian perspective. Int Conf Adv Comput Commun Inform ICACCI. 2015. https://​
doi.​org/​10.​1109/​ICACCI.​2015.​72757​07.
95. Hari SS, Sivakumar MM, Renuga DP, Karthikeyan S, Suriya S. Plant disease detection by leaf image classification
using convolutional neural network. Comput Sci Commun Inf. 2019. https://​doi.​org/​10.​1007/​978-​981-​15-​1384-8_​
16.
96. Jenifa A, Ramalakshmi R, Ramachandran V. Cotton leaf disease classification using deep convolution neural
network for sustainable cotton production. 2019 Int Conf Clean Energy Energy Effic Electron Circuit Sustain Dev
INCCES 2019. 2019. https://​doi.​org/​10.​1109/​INCCE​S47820.​2019.​91677​15.
97. Anand R, Veni S, Aravinth J. An application of image processing techniques for detection of diseases on brinjal
leaves using k-means clustering method. Int Conf Recent Trends Inf Technol ICRTIT. 2016. https://​doi.​org/​10.​1109/​
ICRTIT.​2016.​75695​31.
98. Durmus H, Gunes EO, Kirci M. Disease detection on the leaves of the tomato plants by using deep learning agro-
geoinformatics. Conf Agro-Geoinformatics. 2027. https://​doi.​org/​10.​1109/​Agro-​Geoin​forma​tics.​2017.​80470​16.
99. Nithish EK, Kaushik M, Prakash P, Ajay R, Veni S. Tomato leaf disease detection using convolutional neural network
with data augmentation. Proc Int Conf Commun Electron Syst ICCES. 2020. https://​doi.​org/​10.​1109/​ICCES​48766.​
2020.​09138​030.
100. Singh V, Misra AK. Detection of plant leaf diseases using image segmentation and soft computing techniques. Inf
Process Agric. 2017;4(1):41–9. https://​doi.​org/​10.​1016/j.​inpa.​2016.​10.​005.
101. Sun G, Jia X, Geng T. MPlant diseases recognition based on image processing technology. J Electr Comput Eng.
2018. https://​doi.​org/​10.​1109/6.​100908.
102. Chandramouleeswaran S, Senthil Kumar MD, Professor A. Plant infection detection using image processing. Int J
Mod Eng Res. 2018;8:13–6.
103. Ahmad Supian MB, Madzin H, Albahari E. Plant disease detection and classification using image processing
techniques: a review. Proc 2019 2nd Int Conf Appl Eng ICAE 2019. 2019. https://​doi.​org/​10.​1109/​ICAE4​7758.​2019.​
92217​12.
104. Halder M, Sarkar A, Bahar H. Plant disease detection by image processing: a literature review. SDRP J Food Sci
Technol. 2018;3(6):534–8. https://​doi.​org/​10.​25177/​jfst.3.​6.6.
105. Ghosal S, Sarkar K. Rice leaf diseases classification using CNN with transfer learning. IEEE Calcutta Conf CALCON.
2020. https://​doi.​org/​10.​1109/​CALCO​N49167.​2020.​91064​23.
106. Devaraj A, Rathan K, Jaahnavi S, Indira K. Identification of plant disease using image processing technique. Proc
2019 IEEE Int Conf Commun Signal Process ICCSP 2019. 2019. https://​doi.​org/​10.​1109/​ICCSP.​2019.​86980​56.
107. Suresh M, Gopinath D, Hemavarthini M, Jayanthan K, Krishnan M. Plant disease detection using image processing.
Int J Eng Res Technol. 2022. https://​doi.​org/​10.​1007/​978-​981-​15-​4032-5_​83.
108. Iqbal MA, Talukder KH. Detection of potato disease using image segmentation and machine learning 2020. Int
Conf Wirel Commun Signal Process Networking. 2020. https://​doi.​org/​10.​1109/​WiSPN​ET486​89.​2020.​91985​63.
109. Ali-Al-Alvy M, Khan GK, Alam MJ, Islam S, Rahman M, Rahman MS. Rose plant disease detection using deep learn-
ing. Conf Trends Electron Inform. 2023;3(03):1244–9. https://​doi.​org/​10.​1109/​ICOEI​56765.​2023.​10126​031.
110. Howlader MR, Habiba U, Faisal RH, Rahman MM. “automatic recognition of guava leaf diseases using deep convo-
lution neural network”, ­2nd Int. Conf Electr Comput Commun Eng ECCE. 2019;2019:1–5. https://​doi.​org/​10.​1109/​
ECACE.​2019.​86794​21.
111. Bhagat Patil AR, et al. A literature review on detection of plant diseases. Eur J Mol Clin Med. 2020;7(07):1606.
112. Mohanty SP, Hughes DP, Salathé M. Using deep learning for image-based plant disease detection. Front Plant Sci.
2016;7:1–10. https://​doi.​org/​10.​3389/​fpls.​2016.​01419.
113. Ranjan M, Rajiv Weginwar M, Joshi N, Ingole A. Detection and classification of leaf disease using artificial neural
network. Int J Tech Res Appl. 2015;3(3):331–3.
114. Ramya V, Lydia MA. leaf disease detection and classification using neural networks. Int J Adv Res Comput Com-
mun Eng. 2016;5(11):207–10. https://​doi.​org/​10.​17148/​IJARC​CE.​2016.​51144.
115. Patil R, Gulvani S. plant disease detection using neural network : a review. J Emerg Technol Innov Res.
2019;6(2):151–5.
Demilie J ournal of Big Data (2024) 11:5 Page 42 of 43

116. Amjad K, Ghous H. Critical review on multi-crops leaves disease detection using artificial intelligence methods. Int
J Sci Eng Res. 2021;12:2.
117. Singh MK, Chetia S. Detection and classification of plant leaf diseases in image processing using MATLAB. Int J Life
Sci Res. 2017;5(4):120–4.
118. Raut S, Ingole K. Review on fruit disease detection using image processing techniques. Int Res J Eng Technol.
2017;4(4):22–4.
119. Ramakrishnan M, Sahaya ANA. “Groundnut leaf disease detection and classification by using back probagation
algorithm”, 2015 Int. Conf Commun Signal Process ICCSP. 2015;2015:964–8. https://​doi.​org/​10.​1109/​ICCSP.​2015.​
73226​41.
120. Ngugi LC, Abelwahab M, Abo-Zahhad M. Recent advances in image processing techniques for automated leaf
pest and disease recognition—a review. Inf Process Agric. 2021;8(1):27–51. https://​doi.​org/​10.​1016/j.​inpa.​2020.​04.​
004.
121. Ishak S, Rahiman MHF, Kanafiah SNAM, Saad H. Leaf disease classification using artificial neural network. J Teknol.
2015;77(17):109–14. https://​doi.​org/​10.​11113/​jt.​v77.​6463.
122. Hrishikesh KP, Lokhande SS. detection and classification of plant leaf diseases using ANN. Int J Sci Eng Res.
2013;4(8):1777–80.
123. Sachdeva G, Singh P, Kaur P. Plant leaf disease classification using deep convolutional neural network with bayes-
ian learning. Mater Today Proc. 2021;45:5584–90. https://​doi.​org/​10.​1016/j.​matpr.​2021.​02.​312.
124. Kaur P, Harnal S, Tiwari R, Upadhyay S, Bhatia S, Mashat A. Recognition of leaf disease using hybrid convolutional
neural network by applying feature reduction. Sensors. 2022. https://​doi.​org/​10.​3390/​s2202​0575.
125. Kiani E, Mamedov T. Identification of plant disease infection using soft-computing: application to modern botany.
Procedia Comput Sci. 2017;120:893–900. https://​doi.​org/​10.​1016/j.​procs.​2017.​11.​323.
126. Z. Chuanlei, Z. Shanwen, Y. Jucheng, S. Yancui, and C. Jia, “Apple leaf disease identification using genetic algorithm
and correlation based feature selection method.” 2017.
127. Patil NS. Identification of paddy leaf diseases using evolutionary and machine learning methods. Turkish J Comput
Math Educ. 2021;12(2):1672–86. https://​doi.​org/​10.​17762/​turco​mat.​v12i2.​1503.
128. S. Jasrotiaa, J. Yadava, N. Rajpala, M. Arorab, and J. Chaudhary, “Convolutional Neural Network Based Maize Plant
Disease Identification.” pp. 1712–1721, 2023.
129. Rekha ASRMG. Detection of unhealthy plant leaves using image processing with arduin. CIKITUSI J Multidiscip Res.
2019;6(3):163–7. https://​doi.​org/​10.​35940/​ijitee.​b1233.​1292s​419.
130. Nagamani HS, Saroja DH. Research methods for plant health detection using computer vision techniques: a
survey. Int J Appl Eng Res. 2019;14(7):1627. https://​doi.​org/​10.​3622/​ijaer/​14.7.​2019.​1627-​1632.
131. Hassan RJ, Abdulazeez AM. Plant leaf disease detection by using different classification techniques comparative.
Asian J Res Comput Sci. 2021. https://​doi.​org/​10.​9734/​ajrcos/​2021/​v8i43​0205.
132. Mohanapriya K, Balasubramani M. Recognition of unhealthy plant leaves using naive bayes classifier. IOP Conf Ser
Mater Sci Eng. 2016. https://​doi.​org/​10.​1088/​1757-​899X/​561/1/​012094.
133. FRF Padao EA Maravillas. 2026. Using Naïve Bayesian method for plant leaf classification based on shape and
texture features Humanoid, Nanotechnology. Inf Technol Commun Control Environ Manag Conf. https://​doi.​org/​
10.​1109/​HNICEM.​2015.​73931​79
134. Sharma R, et al. Plant disease diagnosis and image classification using deep learning. Comput Mater Contin.
2022;71(2):2125–40. https://​doi.​org/​10.​32604/​cmc.​2022.​020017.
135. Prof.D.D.Pukale, G. Kokru, S. Nadar, S. Dhar, and S. Singh. 2019. A disease prediction and rectification system for
banana plant. 6, no. 5, pp. 535–539,.
136. FarhanAlHaque ASM, Hafiz R, Hakim MA, RasiqulIslam GM. Computer vision system for guava disease detec-
tion and recommend curative solution using deep learning approach. Int Conf Comput Inf Technol ICCIT. 2019.
https://​doi.​org/​10.​1109/​ICCIT​48885.​2019.​90385​98.
137. Singh A, Kaur H. Potato plant leaves disease detection and classification using machine learning methodologies.
IOP Conf Ser Mater Sci Eng. 2021;1022:1–10. https://​doi.​org/​10.​1088/​1757-​899X/​1022/1/​012121.
138. Elfatimi E, Eryigit R, Elfatimi L. Beans leaf diseases classification using mobilenet models. IEEE Access.
2022;10:9471–82. https://​doi.​org/​10.​1109/​ACCESS.​2022.​31428​17.
139. Zhang X, Qiao Y, Meng F, Fan C, Zhang M. Identification of maize leaf diseases using improved deep convolutional
neural networks. IEEE Access. 2018;6:30370–7. https://​doi.​org/​10.​1109/​ACCESS.​2018.​28444​05.
140. Zaki MA, Narejo S, Ahsan M, Zai S, Anjum MR, Din NU. Image-based onion disease (purple blotch) detection using
deep convolutional neural network. Int J Adv Comput Sci Appl. 2021;12(5):448–58. https://​doi.​org/​10.​14569/​
IJACSA.​2021.​01205​56.
141. Bal F, Kayaalp F. Review of machine learning and deep learning models in agriculture. Int Adv Res Eng J.
2021;5(2):309–23. https://​doi.​org/​10.​35860/​iarej.​848458.
142. Wang Y, Wang H, Peng Z. Rice diseases detection and classification using attention based neural network and
bayesian optimization. Expert Syst Appl. 2021;178:1–11. https://​doi.​org/​10.​1016/j.​eswa.​2021.​114770.
143. M. Govardhan and V. M. B, “Diagnosis of tomato plant diseases using random forest.” 2019.
144. Devi N, et al. Categorizing diseases from leaf images using a hybrid learning model. Symmetry. 2021;13(11):1–14.
https://​doi.​org/​10.​3390/​sym13​112073.
145. Thilagavathi K, Kavitha K, Praba RD, Arina SVAJ, Sahana RC. Detection of diseases in sugarcane using image pro-
cessing techniques. Biosci Biotechnol Res Commun. 2020;13(11):109–15. https://​doi.​org/​10.​21786/​bbrc/​13.​11/​24.
146. Aduwo JR, Mwebaze E, Quinn JA. “Automated vision-based diagnosis of cassava mosaic disease. Proc ICDM Work
Data Min Agric. 2010;1:1–9.
147. Lamba M, Gigras Y, Dhull A. Classification of plant diseases using machine and deep learning. Open Comput Sci.
2021;11(1):491–508. https://​doi.​org/​10.​1515/​comp-​2020-​0122.
148. Khan F, Zafar N, Tahir MN, Aqib M, Waheed H, Haroon Z. A mobile-based system for maize plant leaf disease detec-
tion and classification using deep learning. Front Plant Sci. 2023;14(May):1–18. https://​doi.​org/​10.​3389/​fpls.​2023.​
10793​66.
Demilie Journal of Big Data (2024) 11:5 Page 43 of 43

149. Rekha V, Reddy LV, Chaudhari SV, Gopi A, Nithiya C, Ahamed SK. Automated deep learning with wavelet neural
network based rice plant classification internet things. Proc Technol. 2023. https://​doi.​org/​10.​1109/​IDCIo​T56793.​
2023.​10053​487.
150. Hasan RI, Yusuf SM, Mohd Rahim MS, Alzubaidi L. Automatic clustering and classification of coffee leaf diseases
based on an extended kernel density estimation approach. Plants. 2023. https://​doi.​org/​10.​3390/​plant​s1208​1603.
151. Kirti N. A multi-crop disease identification approach based on residual attention learning. J Intell Syst. 2023.
https://​doi.​org/​10.​1515/​jisys-​2022-​0248.
152. Ullah N, Khan JA, Almakdi S, El-rashidy N, El-sappagh S, Ali F. An effective approach for plant leaf diseases classi fi
cation based on a novel deepplantnet deep learning model. Plant Sci Front. 2023. https://​doi.​org/​10.​3389/​fpls.​
2023.​12127​47.
153. Alghamdi H, Turki T. PDD-Net: plant disease diagnoses using multilevel and multiscale convolutional neural
network features. Agric. 2023. https://​doi.​org/​10.​3390/​agric​ultur​e1305​1072.
154. Ulutaş H, Aslantaş V. Design of efficient methods for the detection of tomato leaf disease utilizing proposed
ensemble CNN model. Electron. 2023. https://​doi.​org/​10.​3390/​elect​ronic​s1204​0827.
155. Bajpai C, Sahu R, Naik KJ. Deep learning model for plant-leaf disease detection in precision agriculture. IJISTA.
2023. https://​doi.​org/​10.​1504/​IJISTA.​2023.​130562.
156. Ahmed M, Ahmed A. Palm tree disease detection and classification using residual network and transfer learning of
inception ResNet. PLoS ONE. 2023. https://​doi.​org/​10.​1371/​journ​al.​pone.​02822​50.
157. Chauhan D, Walia R, Singh C, Deivakani M, Kumbhkar M. Detection of maize disease using random forest classifica-
tion algorithm. Turkish J Comput Math Educ. 2021;12(9):715–20.
158. Jadon S. SSM-Net for plants disease identification in low data regime. IEEE ITU Int Conf Artif Intell Good. 2020.
https://​doi.​org/​10.​1109/​AI4G5​0087.​2020.​93110​73.
159. Mugithe PK, Mudunuri RV, Rajasekar B, Karthikeyan S. Image processing technique for automatic detection of
plant diseases and alerting system in agricultural farms ICCSP IEEE. Int Conf Commun Signal Process. 2020. https://​
doi.​org/​10.​1109/​ICCSP​48568.​2020.​91820​65.
160. Ganguly SSN. A survey on image processing techniques for glaucoma detection. Int J Eng Res Technol.
2016;4(29):1–5.
161. Simhadri CG, Kondaveeti HK. Automatic recognition of rice leaf diseases using transfer learning. Agronomy.
2023;13(4):1–25. https://​doi.​org/​10.​3390/​agron​omy13​040961.
162. Jackulin C, Murugavalli S. “A comprehensive review on detection of plant disease using machine learning and
deep learning approaches. Meas Sensors. 2022;24:100441. https://​doi.​org/​10.​1016/j.​measen.​2022.​100441.
163. Jung M, et al. Construction of deep learning-based disease detection model in plants. Sci Rep. 2023;13(1):1–13.
https://​doi.​org/​10.​1038/​s41598-​023-​34549-2.
164. Hari P, Singh MP. A lightweight convolutional neural network for disease detection of fruit leaves. Neural Comput
Appl. 2023;35(20):14855–66. https://​doi.​org/​10.​1007/​s00521-​023-​08496-y.
165. Mohammed L, Yusoff Y. Detection and classification of plant leaf diseases using digtal image processing methods:
a review. ASEAN Eng J. 2023;13(1):1–9. https://​doi.​org/​10.​11113/​aej.​V13.​17460.
166. Tholkapiyan M, Devi BA, Bhatt D, Kumar ES, Kirubakaran S, Kumar R. Performance analysis of rice plant dis-
eases identification and classification methodology. Wireless Pers Commun. 2023. https://​doi.​org/​10.​1007/​
s11277-​023-​10333-3.
167. Li L, Zhang S, Wang B. “Plant disease detection and classification by deep learning—a review. IEEE Access.
2021;9:56683–98. https://​doi.​org/​10.​1109/​ACCESS.​2021.​30696​46.
168. Nguyen C, Sagan V, Maimaitiyiming M, Maimaitijiang M, Bhadra S, Kwasniewski MT. Early detection of plant viral
disease using hyperspectral imaging and deep learning. Sensors. 2021;21(3):1–23. https://​doi.​org/​10.​3390/​s2103​
0742.

Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.