AI-Powered Disease Detection and Remedy

Recommendation System for the Colocasia Plant

Annie Silviya S H Sriman B Manyali Hassan Kaif
Department of Computer Science and Department of Computer Science and Department of Computer Science and
Engineering Engineering Engineering
Rajalakshmi Institute of Technology Vel Tech Rangarajan Dr.Sagunthala Rajalakshmi Institute of Technology
Chennai R&D Institute of Science and Chennai
[email protected] Technology manyalihassankaif.2022.cse@ritchenn
Chennai ai.edu.in
[email protected]

Lavanya R Nishanthini S Mohammed Muhsin A

Department of Computer Science and Department of Computer Science and Department of Computer Science and
Engineering Engineering Engineering
Rajalakshmi Institute of Technology Rajalakshmi Institute of Technology Rajalakshmi Institute of Technology
Chennai Chennai Chennai
[email protected] [email protected] Mohammedmuhsin.a.2022.cse@ritche
.in nnai.edu.in

Abstract—The ever-present need for super-efficient and pose severe threats to taro cultivation, causing substantial
nutrition-packed food has always challenged the agricultural yield losses and financial burdens on farmers. Leaf blight,
industry. Diseases and pests lead the charge towards making this caused by Phytophthora colocasiae, is a highly destructive
issue worse with their ever-growing impact. Crop quality and disease that can reduce taro yield by 25-50%, making
yield both suffer immensely. One such crop struggling with this
cultivation unprofitable.
condition is Taro, better known by its scientific name, Colocasia
esculenta. It is a tropical vegetable grown in several countries in
Asia and Africa. Like many others, Taro suffers from pestilence It infects at a fast rate in hot and moist conditions, forming
and diseases, and the damage can be severe due to its versatility lesions, wilting, and early defoliation. The mosaic virus, on
and ingredients as well as its nutritional profile, which consists the other hand, is mostly spread by aphids and causes mottled,
of a ton of carbohydrates and fibre while having very low deformed, or curled leaves, reducing plant strength and
calories. Cultivating taro is not economically viable for most due eventually productivity [14]. Left uncontrolled, such diseases
to Phytophthora colocasiae, which brings forth “leaf blight.” can damage entire fields, impacting small-scale farmers as
This condition reduces the yield by approximately 25%-50%. well as huge-scale agricultural businesses. Conventional
The plant is also affected by a mosaic virus, which is transmitted
mainly by aphids. This virus weakens the plant and also destroys
disease detection techniques are based totally on manual
leaves. To ensure the impacts are not overly excessive, prompt exams, which are time-consuming, hard work in-depth, and
action towards disease control is imperative. AI solutions such prone to human blunders. These conventional techniques tend
as the AI MobileNetV2, EfficientNet, and ConvNeXt, which to bring about late diagnoses, giving sicknesses adequate time
assist in recognizing blight leaf and marking infection of taro to spread before corrective action can be taken. To conquer
leaves, will be implemented in this study. Alongside image those obstacles, the incorporation of artificial intelligence
classification, actionable suggestion systems built using the (AI) and machine learning to know (ML) [20] technologies
Hugging Face library's Transformer-based models will provide into plant ailment detection systems offers a
immediate and useful suggestions, helping taro leaf disease possible solution. In plant health analysis, disease symptom
management be effortless and efficient.
identification, and early alert provision to farmers, deep
Keywords—Agriculture, grid system, AI models, efficiency, learning image classification models have shown remarkable
resistance. success.

I. INTRODUCTION Using computer vision-based [5] disease detection, mass

agricultural surveillance finds a cheap and scalable solution
Agriculture is the cornerstone of maintaining human that lessens reliance on professional services. In this work,
existence through the provision of an escalating food image classification of taro leaf images and symptoms of leaf
production demand. The industry, though, is confronted by blight and mosaic virus detection at high accuracy [4] rates
various challenges, especially crop diseases that have using MobileNetV2 [15], EfficientNet, and ConvNeXt
profound effects on yield, quality, and economic profitability. models. Using convolutional neural networks (CNNs), these
Among such crops by such challenges is taro (Colocasia deep learning [3] models extract intricate patterns and
esculenta), a common crop in most tropical and subtropical separate healthy from infected plants. Beyond the diagnosis
countries of Asia, Africa, and the Pacific Islands. Taro is of diseases, efficient disease management relies heavily on
widely cultivated for its edible corms and leaves, which are timely and exact recommendations. This study utilizes the
rich in carbohydrates, fiber, and essential nutrients. Despite AI-driven recommendation system with transformer-based
its significance, diseases such as leaf blight and mosaic virus [16] Hugging Face models, namely FLAN T5 [11] as the
analysis and improvement model and BART as the methods, including transfer learning [21] utilizing pre-trained
summarizing model. After disease detection, the system EfficientNet, ConvNeXt, and MobileNetV2 [15], have been
generates the human-friendly text output summarizing the embraced for enhancing the detection accuracy [4].
most effective strategies and precautionary measures. Recommendation systems were also incorporated alongside
disease classification into agricultural research. Certain
studies have used natural language processing (NLP) models,
including BERT and GPT, to make textual recommendations
on managing disease.

These methodologies are based on substantial fine-tuning and

might not necessarily generalize to all crops and conditions.
For better interpretability and efficiency, this study utilizes
transformer-based [16] models from the Hugging Face
library, such as FLAN-T5 [11] for reasoning and BART for
abstractive summarization, to provide clear, concise, and
effective solutions for diagnosed diseases. The integration of
deep learning [3] for image classification and transformers
for recommendation generation offers a strong solution for
early disease detection in taro plants. Although previous work
has shown that these individual technologies work
individually, their combination into one pipeline for detection
Fig. 1. Pipeline for Disease Detection and and remedy generation is still not very well explored. This
Recommendation system work extends the existing research by adopting a hybrid
system that not only detects mosaic virus and leaf blight in
This makes the information implementable and intelligible taro leaves but also offers human-readable solutions, closing
for farmers, reducing the need for deep research and reliance the loop between AI-based disease diagnosis and actionable
on agricultural specialists. To provide efficient agricultural recommendations.
recommendations for maintaining healthy crop yields and
preventing diseases, the system uses deep learning [3] and III. METHODOLOGY
natural language processing (NLP). This study proves how
AI-generated remedies have the capability to enhance This study uses deep learning [3] to detect and classify the
agriculture in modern days. This system improves precision Colocasia (taro) leaf diseases like Mosaic Virus and Leaf
agriculture, lowers manual labour, and advances sustainable Blight. The crucial work for the primary classification model
agriculture by merging image-based disease detection with like MobileNetV2, EfficientNet, and ConvNeXt is to identify
intelligent recommendations. It provides the farmers access the plant disease in more accurate and efficient manner. The
to the latest, data-supported information so they can avoid pre-processing techniques, like false positives, normalization,
crop diseases before they become serious problems. Hence, and augmentation, were performed by dataset to ensure the
by assuring that vital crops like taro persist as a feasible food model's stability and reliability. The trained model undergoes
source for millions of people, the suggested remedy not only various sets of image samples to improve the generalization
enhances the yield and quality but also tackles global and to decrease both false positives and false negatives. The
food security [1]. deep learning model plays a crucial role for prompt action and
efficient disease prevention. The inclusion of a deep learning
II. RELATED WORK model enables the accurate and swift identification of
infected Colocasia (taro) leaves.
Artificial Intelligence and deep learning [3] have seen
substantial growth in precision agriculture in recent years, Beyond the classification, this research encompasses a
especially in detecting plant disease [1]. Various researchers Transformer-based [16] recommendation system to offer
have utilized convolutional neural networks (CNNs) [8] to practical assistance for farmers. This system leverages pre-
identify plant diseases [1] from leaf images. Researchers have trained language models from the Hugging Face library,
used models like VGG16, ResNet, and MobileNet, which including Facebook BART Large CNN for summarizing
have proved effective in identifying diseased and healthy treatment recommendations and Google Flan T5 [11] Large
plants with high accuracy [4]. Although these models for upgrading disease descriptions. By merging deep
successfully pick features from images, their effectiveness learning-powered detection with NLP-driven
and computational expense differ for their suitability in recommendations, farmers can receive evident, concise, and
practical agricultural applications. Some efforts have contextually relevant assistance through an intuitive and user-
explicitly been devoted to taro plant disease detection. friendly interface. This approach supports early disease
identification, showcases sustainable farming methods, and
Under the traditional method for disease detection, manual advances precision agriculture by integrating state-of-the-art
inspection was the practice, which was time-consuming and image recognition [7] with innovative natural language
susceptible to erroneous results because symptom variations processing techniques.
across various environments were a problem. Subsequent
studies employed more advanced machine learning [20] IV. MODEL ARCHITECTURE
classifiers, including support vector machines (SVMs) and
random forests, but suffer from scalability [6] and flexibility 1. Disease Detection System
since handcrafted feature extraction [13] is still required. For In an organized procedure, the first step of the proposed
overcoming these problems, contemporary deep learning [3] model for disease detection is the input layer. It includes
resizing the images to 224 x 224 x 3 pixel size. The resizing
of images helps to attain equality in image size and colour
channels, which facilitates the model to ideally manage the
input data. Following, the deep learning models like
MobileNetv2, EfficientNet, and ConvNeXt contain a feature
extraction [13] layer to process the resized images. In order
to collect a leaf's details, these models analyze remarkable
patterns such as colour change, texture, and anomalies caused
by disease. The extracted feature portrays the visual
properties of the leaf and is stored in a high-dimensional
feature map that is 10 x 10 x 1536. Then pooling layer is used
to compress the extracted feature maps to 7 x 7 x 1536 for
optimized computation efficiency. This dimensional
minimization enhances the model capacities.

Fig. 3. Architecture diagram of the detecting system

Fig. 2. Hyperparameter Comparison of Deep 2. Recommendation system
Learning Models
The recommendation system is responsible for helping
The dropout layer in the training process disables neurons farmers manage diseases by offering actionable information
randomly to allow the model to learn strong features. This based on the classification outcome of Colocasia leaf
part of the process enhances the generalization and controls
diseases. Following the deep learning [3] model categories, if
overfitting. Then, the extracted features are passed to the fully
connected layer. In there, the layers are where the patterns a leaf is damaged by Leaf Blight or a mosaic virus, the system
and correlations are pooled. This makes the model understand provides accurate and simple disease prevention and
disease-related objectives. Another dropout layer is added at treatment remedies. This AI-based outcome reduces
this point for engaging things consistently and preventing the dependence on a masterful approach and provides farmers
model not to overfitting in training. At last, the SoftMax layer with immediate advice to neutralize the impact of these
is used to transform the learned pattern into a probability diseases. Using Transformer-based [16] models like Google
distribution. The model can categorize the input image into
FLAN-T5 [11] Large and Facebook BART Large CNN, the
categorized classes. The output has various disease classes
dependent on the surface of the leaves, whether Leaf Blight system enhances the availability of agricultural information
Dorsal, Leaf Blight Ventral, Mosaic Virus Dorsal, or Mosaic by paraphrasing and summarizing detailed data into farmer-
Virus Ventral. speak. The recommendation system can transform
classification output into a knowledge base for best training
The model predicts healthy surface leaves from both the in agricultural control measures by converting it into a pre-
dorsal and ventral sides. This helps in efficient disease
defined knowledge base.
management and accuracy in the identification of the disease
type for initial detection. Due to the accuracy of the results, This expertise base is collected from agricultural research
the model helps in the detection of infection at an initial stage
articles, government reports, and professional assistance to
before causing huge losses. To support the disease detection
model, an AI-based recommendation system is incorporated certify the credibility of the remedies provided. When a
to offer actionable information for disease management. disease is detected, the system provides a response that
Utilizing Hugging Face Transformer-based [16] models, e.g., consists of a clarification of the infection, its symptoms, and
Google FLAN-T5 [11] Large for rewording and Facebook the best treatment preferences. For example, if a leaf is
BART Large CNN for summarizing, the system provides affected with Leaf Blight, the system can recommend a
readable, understandable suggestions that are specifically
fungicide application, best irrigation methods, and crop
made for farmers. These suggestions involve infection-
specific treatment processes, appropriate pesticides, and rotation methods to exclude further infection. In the same
prophylactic methods. way, recommendations for the Mosaic Virus might focus on
pest control measures to limit aphid and virus-resistant plant time diagnosis and intervention. This lag in disease detection
cultivar transmission. The recommendation system uses AI- resulted in massive crop losses, which impacted the farmer's
based text generation to resolve technical facts without livelihood and the food supply chain. Artificial intelligence
sacrificing accuracy for consistency and usability [4]. The (AI) in agriculture reveals a workable way to address these
FLAN-T5 from Google [11]. To make the insight available to problems.
farmers of all ability tiers, a large model is used to summarize
Based on images, the deep learning algorithms have
complex agricultural terminology into an evident format.
illustrated notable efficacy in diagnosing diseases and
However, text summarization is done by the Facebook BART offering high precision. Disease can be identified initially
Large CNN model [9], which summarizes lengthy with AI-driven image classification methods, allowing for
explanations into fragments. This mixture makes it viable for prompt and efficient treatment.
the recommendation system to provide farmers evident and
The transformer-based [16] recommendation system
clear assistance on how to handle diseases. The system urges
provides farmers access to treatment plans and practical
proactive measures to lower future infection in addition to
insights in a user-friendly interface. The need to create a
offering disease treatment. The farmers are given suggestions
readily available, self-operating, and intelligent system that
on integrated pest control techniques, appropriate fertilization
can detect crop disease and offer farmers correct suggestions,
frameworks, and soil health management to improve the
which will ultimately lead to improved agricultural
resistance of their plants. Moreover, new research and the
sustainability and efficiency, is what motivates this research
latest data can be added to the system consistently, ensuring
[14].
that the suggestions remain relevant and useful.

The system not only upgrades early disease detection but also VI. PROPOSED SOLUTION
provides farmers with the idea they need to implement eco- 1. MobileNetV2
friendly farming techniques by merging AI-enabled
suggestions with deep learning-based disease detection. By It is one of the main deep learning models in this project for
applying cutting-edge farming strategies in real-world disease diagnosing in Colocasia (taro) leaves. The ultimate
scenarios for those who need it most, this integrated method goal of MobileNetV2 [15] is to detect whether the input
enhances efficiency, lowers harvest losses, and ensures global images are healthy or diseased by extracting the beneficial
food security [1]. features from them. MobileNetv2 [15] is especially well-
fitted for this use since it is productive and lightweight in
design, making it unique for deployment in modern
agricultural applications like mobile apps and edge devices.

The model takes images rescaled to 224 × 224 × 3 pixels as

input for work at the feature extraction [13] stage. The depth-
wise separable convolutions and inverted residual blocks of
MobileNetV2 [15] effectively extract salient patterns,
including texture changes, discolorations, and lesions
characteristic of Leaf Blight and Mosaic Virus. These
extracted features are subsequently sent to a fully connected
layer and a SoftMax classifier to predict the occurrence and
kind of disease. Leveraging MobileNetV2's [15] low
computational overhead, the system makes it possible for
quick, accurate, and lightweight disease detection even within
the power constraints of agricultural settings.

Fig. 4. Recommendation system

V. MOTIVATION

Agriculture is still the pillar of food security [1], but it

continues to be beset by challenges from crop diseases, which
impact both yield and quality. Taro (Colocasia esculenta), an
important staple crop in most areas, is especially susceptible
to diseases like Leaf Blight and Mosaic Virus. Conventional
disease detection involves manual observation, which is time-
consuming, labor-intensive, and subject to errors based on
human subjectivity. The farmers in rural or resource-poor
areas, who do not always have a direct connection to
agricultural experts, are facing the issue of unachievable on- Fig. 5. Confusion matrix of MobileNetV2
2. EfficientNet 3. ConvNeXt

EfficientNet is used as a feature extractor for Colocasia (taro) ConvNeXt is a key element of the system, which is a feature
leaf disease detection [18] in the system proposed. It is extractor for the disease classification of Colocasia (taro)
scalable and optimized in nature and thus can maintain high leaves from diseases like Leaf Blight and Mosaic Virus. The
accuracy [4] with efficient computations. EfficientNet is other deep learning creations, like enhanced activation
different from the traditional convolutional neural network functions, depth-wise convolutions, and Layer Norm,
[8] because it utilizes compound scaling to change depth, ConvNeXt improves the traditional Convolutional Neural
width, and resolution equally for better model performance Network (CNN) [8] architecture, gaining inspiration from the
with minimal growth in computational requirements. recent advancement in transformer-based [6] models.
Appreciation of these innovations can efficiently process
In this framework, EfficientNet takes resized input images of
images and extract detailed data.
224 × 224 × 3 pixels and derives significant disease-related
features, including lesions, coloration changes, and texture ConvNeXt uses this architecture to retrieve essential disease-
changes. The sophisticated convolutional layers of the model related details from input images with pixel sizes of 224 ×
detect subtle patterns that distinguish healthy leaves from 224 × 3, including texture anomalies, color aberrations, and
infected ones. Following that, these extracted features are sore morphologies. Its hierarchical structure, which
passed through fully connected layers and a SoftMax demonstrates dimensional pattern recognition, makes it
classifier to categorize the leaf as leaf Blight, Mosaic Virus effective at drawing precise distinctions between healthy and
or healthy. For the proficiency between efficiency and diseased leaves. To classify the leaf into appropriate classes,
accuracy [4], EfficientNet is the perfect option for real-time the learned details or features are then fed into a fully
disease detection. connected layer and a SoftMax classifier. Due to its powerful
function extraction [13] capability, the ConvNeXt verifies
precise disease detection with ideal performance adapted for
real-time use in agriculture.

Fig. 6. EfficientNet’s Validation loss and accuracy

Fig. 8. Training results of ConvNeXt

4. BART (facebook/bart-large-cnn)

BART, a complete learning transformer model that generates

the recommendation system's crisp and structured
summaries. Once the disease is identified by the deep
learning models, the data is retrieved from the vast quantity
of data about the specific detected ailments. This includes the
symptoms, severity, and treatment methods.

To make this process easier, BART breaks down the received

content into logical outlines by removing the unnecessary
details. The summarization process of this model ensures that
farmers can obtain suggestions that are practical and easy to
understand. By summarizing complex information and vast
amounts of text into knowledge, this model makes disease
Fig. 7. Comparison of EfficientNet and MobileNetV2 management easier. Further to improve readability, the
finalized summary moves to FLAN-T5 [11] to make it more farmers to make conclusion hastily by avoiding the need for
readable. specialist interaction and human inspection.

5. FLAN-T5

FLAN-T5 [11] enhances the recommendation system by

making disease management assistance more understandable
and readable for farmers. When the disease in Colocasia
(taro) plants is detected, the system generates a primary
response with medication and defensive measures. The
untreated outcomes could be difficult for the user to
understand because of their unstructured manner.

FLAN-T5 achieves this by reorganizing and rephrasing the

generated text into a more understandable and cohesive
format. While protecting important information through Fig. 9. Screenshot of Colocasia Disease Detection
instruction tuning, it makes suggestions more user-friendly. Interface
This enables the understanding of significant disease
management approaches by farmers with varying levels of
literacy. FLAN-T5's [11] capability to generate human-like
responses reduces the need for agricultural experts and
enables prompt intervention. The model reduces crop losses
through detection, rapid advice, and promotion of sustainable
agriculture and enables efficient disease control through
clarified communication.

VII. RESULT AND DISCUSSION

Comparing the tested systems, which included MobilenetV2,

EfficientNet, and ConvNeXt, we obtained excellent accuracy
in detecting healthy or diseased-colonized parts. For Fig. 10. Screenshot of Model Prediction Result
processing, crucial preconditions like resizing, normalization,
and augmentation are utilized from a dataset. To prepare it for
modelling to make it vigorous consisted of Colocasia (taro)
leaf images that either had Mosaic Virus or Leaf Blight.
ConvNeXt utilizes its cutting-edge convolutional architecture
to enhance feature extraction and accuracy. MobileNetV2 is
a lightweight model that provides quick assumption and is
suitable for real-time deployment. EfficientNet offered a
balance between accuracy and computational efficiency.
With an exact solution for early detection and crop loss
prevention, classification output shows the dominance of
deep learning in plant disease diagnosis. The system contains
an AI-driven recommendation module to help farmers with Fig. 11. Screenshot of Remedy Recommendation
practical and realistic recommendations for disease
VIII. FUTURE WORKS
management.
In the future, the futuristic deep learning model can upgrade
When used for summarization, Facebook BART effectively
the efficiency and accuracy of disease identification and the
condenses information about diseases into succinct, easily
recommendation system. The vision transformer or self-
comprehensible responses. These ensures that significant
supervised learning can be utilized to enhance the disease
details can be understood by users of all technical skill levels
classification and feature extraction. Another ideal
without being hampered by irrelevant information.
implementation is including image processing with real-time
Additionally, to ensure readability, FLAN-T5 reorganizes
surrounding variables like temperature and humidity. This
these findings into consistent, logical recommendations. The
helps improve the disease recognition to make it more
coordination of these models permits the flow of details and
accurate. We can also increase the reliability of our model by
confirms the system offers valuable, practical assistance for
including a vast amount of Colocasia (taro) leaf images with
disease precaution and recognition. A crucial advancement in
different angles and lighting conditions. Reaching farmers
agriculture technology is the compositions of deep learning
can be improved by creating web-based and mobile
and natural language processing (NLP). The system permits
applications. This improved to allow the farmer to take
pictures of their leaves and provide an instant, real-time, MobileNetV2 Model," 2024 4th International
accurate disease diagnosis and recommendations. Finally, the Conference on Innovative Research in Applied
capacity of the recommendation system could increase with Science, Engineering and Technology (IRASET),
reinforcement learning, ultimately leading to better accuracy FEZ, Morocco, 2024, pp. 1-6, doi:
and more flexibility in recommendations. 10.1109/IRASET60544.2024.10548207.
[6] S. Rajendran and T. M. Swamy, "Improving Cotton
IX. CONCLUSION Leaf Disease Detection with MobileNetV2: An
The system presented in this paper includes deep learning- Analysis in Comparison with MobileNetV3-Small,"
powered image classification with an AI-powered image 2024 International Conference on System,
recommendation system for enhancing disease detection and Computation, Automation and Networking
handling of Colocasia (taro) leaves. Along with the use of (ICSCAN), PUDUCHERRY, India, 2024, pp. 1-5,
advanced models like MobileNetV2, EfficientNet, and doi: 10.1109/ICSCAN62807.2024.10894642.
ConvNeXt, the system presents accuracy in identifying Leaf [7] D. Li and C. Zhang, "Identification of Pests and
Blight and Mosaic Virus with less intervention from humans. Diseases in Greenhouse Rice Based on ConvNeXt-
Incorporation of BART to summarize and FLAN-T5 to T Neural Network," 2024 International Conference
generate refined responses, which are simple, accurate, and on Distributed Computing and Optimization
practical, helps farmers execute effective methods for disease
Techniques (ICDCOT), Bengaluru, India, 2024, pp.
control. This makes not only the detection process easy but
1-7, doi: 10.1109/ICDCOT61034.2024.10515357.
also avails users with understandable and organized
[8] T. Prabahar Godwin James, P. J, V. G, A. K. M S,
recommendations for preventing plant diseases. The system
promotes sustainable agriculture by minimizing the loss of D. R and A. K. J, "Detecting Plant Leaf Diseases
crops and expert advice dependence. With speedy and with Convolutional Neural Networks (CNN) and
accurate analysis, farmers can intervene in good time, Deployment via Hugging Face," 2024 International
reducing extensive damage and enhancing overall harvest. Conference on Power, Energy, Control and
The recommendation system's dynamic nature makes it as Transmission Systems (ICPECTS), Chennai, India,
user-friendly as possible, making it easier for users with less 2024, pp. 1-6, doi:
technical expertise to comprehend and react to suggestions. 10.1109/ICPECTS62210.2024.10780036.
[9] A. M. Ahmed Zeyad and A. Biradar,
X. REFERENCE
"Advancements in the Efficacy of Flan-T5 for
Abstractive Text Summarization: A Multi-Dataset
[1] R. Kumar, S. Gupta, Y. Srivastava, and H.
Evaluation Using ROUGE and BERTScore," 2024
Srivastava, "AI-Assisted Plant Disease Detection,
International Conference on Advancements in
Crop and Fertilizer Recommendation System," 2024
Power, Communication and Intelligent Systems
5th International Conference for Emerging
(APCI), KANNUR, India, 2024, pp. 1-5, doi:
Technology (INCET), Belgaum, India, 2024, pp. 1-
10.1109/APCI61480.2024.10616418.
6, doi: 10.1109/INCET61516.2024.10592955.
[10] A. Harish, G. Prakash, R. R. Nair, V. B. Iyer and A.
[2] C. Shewale, A. Sardeshmukh, P. Shinde, O. Sapkal,
K. M, "Refining LLMs with Reinforcement
and S. Shinde, "Intelligent System for Crop
Learning for Human-Like Text Generation," 2024
Recommendation and Disease Identification," 2024
IEEE International Conference on Electronics,
IEEE 9th International Conference for Convergence
Computing and Communication Technologies
in Technology (I2CT), Pune, India, 2024, pp. 1-5,
(CONECCT), Bangalore, India, 2024, pp. 1-6, doi:
doi: 10.1109/I2CT61223.2024.10544110.
10.1109/CONECCT62155.2024.10677038.
[3] V. C. Ganesh Reddy, V. M. Arul Xavier, and S. S.
[11] S. A, P. R.P, H. V, S. Kumar R and S. P. S,
Shyni, "Plant Disease Detection and Pesticide
"Abstractive Summarization of YouTube Videos
Recommendation Using Deep Learning," 2024 10th
Using LaMini-Flan-T5 LLM," 2024 Second
International Conference on Advanced Computing
International Conference on Advances in
and Communication Systems (ICACCS),
Information Technology (ICAIT), Chikkamagaluru,
Coimbatore, India, 2024, pp. 2543-2547, doi:
Karnataka, India, 2024, pp. 1-5, doi:
10.1109/ICACCS60874.2024.10717268.
10.1109/ICAIT61638.2024.10690747.
[4] A. R. Gadah, S. Isaak, A. -M. H. Y. Saad and N. A.
[12] B. K, S. V, H. K. Lohar, N. G D, V. S and Y. KN,
Mokhtar, "Revolutionizing Plant Health:
"Arecanut Disease Prediction Using MobileNet V2,
Leveraging EfficientNet B3 for Advanced Tomato
ResNet and VGG-16 Framework," 2024 4th
Disease Detection in Agricultural AI," 2024 IEEE
International Conference on Ubiquitous Computing
Symposium on Industrial Electronics &
and Intelligent Information Systems (ICUIS),
Applications (ISIEA), Kuala Lumpur, Malaysia,
Gobichettipalayam, India, 2024, pp. 1972-1977, doi:
2024, pp. 1-6, doi:
10.1109/ICUIS64676.2024.10866077.
10.1109/ISIEA61920.2024.10607355.
[13] A. Suryavanshi, V. Kukreja and S. Hariharan, "From
[5] S. Bouskour, M. H. Zaggaf and L. Bahatti, "Deep
Detection to Severity Estimation: A Unified CNN-
Learning Recognition of Wheat Leaf Disease Using
 SVM Strategy for Taro Leaf Blight Disease," 2023 [22] W. Zhang, T. Lin, H. Wang and L. Song,
3rd International Conference on Technological "ConvNeXt-CNN Based Intelligent Fault Diagnosis
Advancements in Computational Sciences Method for Rotating Equipment," 2022
(ICTACS), Tashkent, Uzbekistan, 2023, pp. 1385- International Conference on Sensing, Measurement
1391, doi: 10.1109/ICTACS59847.2023.10390391. & Data Analytics in the era of Artificial Intelligence
[14] P. Kanaga Priya, T. Vaishnavi, T. Pavithra, R. (ICSMD), Harbin, China, 2022, pp. 1-6, doi:
Sivaranjani, A. Reethika and G. Ramesh Kalyan, 10.1109/ICSMD57530.2022.10058420.
"Optimized Plant Disease Prediction using CNN and [23] J. Thirunavukkarasu, K. J. Oindrilla, M. Sangeetha
Fertilizer Recommendation Engine," 2023 4th and E. Swetha, "An Efficient Deep Learning
International Conference on Electronics and Approach for Plant Disease Detection," 2022
Sustainable Communication Systems (ICESC), International Conference on Innovative Computing,
Coimbatore, India, 2023, pp. 1660-1665, doi: Intelligent Communication and Smart Electrical
10.1109/ICESC57686.2023.10193314. Systems (ICSES), Chennai, India, 2022, pp. 1-7,
[15] T. M. Fahrudin and I. Z. A. Illah, "SkinMate: doi: 10.1109/ICSES55317.2022.9914063.
Mobile-Based Application for Detecting Multi- [24] A. K. S and A. Negi, "A Detection and Classification
Class Skin Diseases Classification Using Pre- of Cotton Leaf Disease Using a Lightweight CNN
Trained MobileNetV2 on CNN Architecture," 2023 Architecture," 2022 Fourth International
IEEE 9th Information Technology International Conference on Emerging Research in Electronics,
Seminar (ITIS), Batu Malang, Indonesia, 2023, pp. Computer Science and Technology (ICERECT),
1-6, doi: 10.1109/ITIS59651.2023.10420370. Mandya, India, 2022, pp. 1-5, doi:
[16] S. Hao et al., "ConvNeXt-ST-AFF: A Novel Skin 10.1109/ICERECT56837.2022.10060246.
Disease Classification Model Based on Fusion of [25] S. H Annie Silviya, S. B, P. B. Shamini, A.
ConvNeXt and Swin Transformer," in IEEE Access, Elangovan, M. A. R and N. V. Keerthana, "Deep
vol. 11, pp. 117460-117473, 2023, doi: Learning based Plant Leaf Disease Detection and
10.1109/ACCESS.2023.3324042. Classification," 2022 4th International Conference
[17] H. Chen and Y. Liao, "The Function of Crop pest on Inventive Research in Computing Applications
and disease identification based on the improved (ICIRCA), Coimbatore, India, 2022, pp. 702-710,
ConvNeXt Network model," 2023 4th International doi: 10.1109/ICIRCA54612.2022.9985548.
Conference on Electronic Communication and
Artificial Intelligence (ICECAI), Guangzhou,
China, 2023, pp. 387-391, doi:
10.1109/ICECAI58670.2023.10176546.
[18] R. Renugadevi, S. Vaishnavi, S. Santhi and S. Pooja,
"Ladies Finger Leaf Disease Detection using CNN,"
2023 2nd International Conference on Applied
Artificial Intelligence and Computing (ICAAIC),
Salem, India, 2023, pp. 595-601, doi:
10.1109/ICAAIC56838.2023.10140976.
[19] G. Priyadharshini and D. R. Judie Dolly,
"Comparative Investigations on Tomato Leaf
Disease Detection and Classification Using CNN,
R-CNN, Fast R-CNN and Faster R-CNN," 2023 9th
International Conference on Advanced Computing
and Communication Systems (ICACCS),
Coimbatore, India, 2023, pp. 1540-1545, doi:
10.1109/ICACCS57279.2023.10112860.
[20] V. Y, N. Billakanti, K. Veeravalli, A. D. R. N and L.
Kota, "Early Detection of Casava Plant Leaf
Diseases using EfficientNet-B0," 2022 IEEE Delhi
Section Conference (DELCON), New Delhi, India,
2022, pp. 1-5, doi:
10.1109/DELCON54057.2022.9753210.
[21] M. Hassam et al., "A Single Stream Modified
MobileNet V2 and Whale Controlled Entropy Based
Optimization Framework for Citrus Fruit Diseases
Recognition," in IEEE Access, vol. 10, pp. 91828-
91839, 2022, doi: 10.1109/ACCESS.2022.3201338.