International Journal of Current Research and Review Research Article

DOI: http://dx.doi.org/10.31782/IJCRR.2020.122028

Potato Plant Disease Detection Using Convolution

Neural Network
IJCRR
Section: Life Pitchai R1, Sharath Kumar G2, Ashutosh Varma D3, Madhu Babu CH4
Sciences
Sci. Journal Impact 1
Associate Professor, Department of Computer Science and Engineering, B.V. Raju Institute of Technology, Narsapur, Telangana, India;
Factor: 6.1 (2018)
ICV: 90.90 (2018)
2,3
Scholar, Department of Computer Science and Engineering, B.V. Raju Institute of Technology, Narsapur, Telangana, India;
4
Professor, Department of Computer Science and Engineering, B.V. Raju Institute of Technology, Narsapur, Telangana, India.
Copyright@IJCRR

ABSTRACT .
Background: In countries like India, whose primary occupation is agriculture faces a huge loss when the crops get affected by
any disease. These diseases attack the crops in various stages and destroy the entire production. Since most of the diseases
are transmitted from one crop to another there is much need to detect the exact type of disease the crop has been affected so
that farmers can take the required steps to the “save the crops” and production. But detecting the kind of disease that a crop has
been affected is very difficult for farmers since there are various kinds of diseases.
Method: There are so many classification techniques, such as k-Nearest Neighbor Classifier, Probabilistic Neural Network, Ge-
netic Algorithm, Support Vector Machine, and Main Component Analysis, Artificial Neural Network, and Fuzzy Logic. It is difficult
to select the best classification method as compared to other methods the system will be more reliable. This article presents a
dissection of various techniques used to find the disease of the plants.
Results and Observation: The developed system is capable of detecting diseases in plants and is also capable of providing
treatments that can be used against them. To improve the health of the plant, we need to deal with it with sufficient knowledge
of the disease and cure. The framework proposed is implemented using python and the google GPU(Graphical Processor Unit)
used provided 80 % accuracy.
Conclusion: The proposed model used a convolution a neural network model based on SSD mobile network for data training.
Some checkpoints are created after training the model. We need to take the last Model Checkpoint to create a file that is used for
testing. A particular file type with the.pb extension is created by using the checkpoint file. The model provided 80% of accuracy.
Key Words: Plant Disease Detection, Neural networks, Genetic algorithm, KNN, PCA, Fuzzy Logi

INTRODUCTION an incredibly long processing period. The image processing

techniques are then used to solve this problem. Health moni-
India is a cultivated nation that relies on farming for approxi- toring and plant disease identification are very relevant to
mately 70% of its population. Farmers have a wide range sustainable farming. Manually tracking plant diseases is very
for the option of different suitable crops and the search for difficult. It needs a large amount of effort, plant disease ex-
effective plant pesticides. Plant disease leads to a substantial perience and an unnecessary time for production. The analy-
decrease in both farm goods quality and quantity. The aim sis of photographs is also used to identify plant diseases.
of agriculture is not just to feed people but also to provide a
major energy source and solution to the worldwide problem Detecting diseases requires measures such as collection, pre-
of global warming. To order to treat and monitor the disease, processing, segmentation of images, extraction of features
the diagnosis of plant disease is very important at an ear- and classifications. Within the following (Section III) the lit-
lier level. Plant disease studies include analyses of visually erature review provides brief details on the different methods
tracked plant trends. used to detect plant diseases.

During the early days, the specialist in this field was able to Motivation
manually track and examine plant diseases, who would no- The leaves are a delicate part of the plant; the assessment of
tice the light colour changes on the plant leaf, but this takes farm harvest is of dynamic importance. The leaf’s texture

Corresponding Author:
Pitchai R, Associate Professor, Department of Computer Science and Engineering, B.V. Raju Institute of Technology, Narsapur, Telangana,
India; Email: [email protected]
ISSN: 2231-2196 (Print) ISSN: 0975-5241 (Online)
Received: 06.07.2020 Revised: 13.08.2020 Accepted: 27.09.2020 Published: 27.10.2020

Int J Cur Res Rev | Vol 12 • Issue 20 • October 2020 152
 Pitchai et al.: Potato plant disease detection using convolution neural network

and colour is the most significant visual property. There- ing coevolutionary neural networks. Raghavendran S et al.
fore it is important to identify leaf diseases to assess farm (2019)13 addressed the categorisation of leaf disease using
products, enhance the market value and comply with quality vector machine support technique. Sankaran et al. (2010)14
standards. This is also important to recognise and take ad- surveyed some of the advanced plant disease identification
ditional measures to further spread diseases. It’s going to be techniques.
too slow. When physical methods are used to classify and
Tyr Wiesner-Hanks et al. (2019)15 outlined a two-step pro-
categorize the method, we also need the professionals to en-
cess for producing large quantities of high-quality training
sure it gets incorrect, and who’s less readily available.
data with limited expert feedback. Liu, LOuyang, W and
The works are categorized according to colour, scale, etc. Wang, X et al. (2020)16 reviewed more than 300 research
When such procedures are documented in an automated sys- papers including several views on the detection of generic
tem by using the correct software design language, the effort objects: frameworks for identification, representation of ob-
becomes faster and error-free. There are two main features to ject characteristics, generation of object proposals, context
be achieved, speed and precision in the methods of machine- modelling, training strategies, and metrics for assessment. M
learning for plant disease detection. Technologies such as Pothen, E and D. M. L. Pai, (2020)17 suggested a classifica-
automated detection and classification of plant diseases us- tion system for the rice leaf disease.
ing floor image processing techniques are required. It proves
The following parameters were considered: bacterial leaf
useful for farmers and warns them at the right time before the
blight, Leaf smut and Brown spot diseased images, and
disease spreads over a wide area.
Otsu’s segmentation method was used. Sun et al. (2020)18
proposed the multi-scale detection of feature fusion using
Literature Survey neural networks of convolution. J. Li et al. (2020)19 proposed
T. Van der Zwet et al. (2002)1 developed a model of identi- a system for the identification, location and study of marine
fication of plant diseases using CNN to classify 13 differ- cucumber using a Faster R-CNN. W. As a new extraction
ent types of plant diseases. Tatem and Rogers et al. (2006)2 design for the Rice Blast detection Chen et al. (2020)20 im-
researched the Global Transportation Networks and the plemented a revolutionary spore germination method. Rasul
transmission of infectious diseases and their effects on epi- Ruzmetov et al. (2020)21 reviewed the morphological prop-
demiology in particular. J. R. Rohr et al. (2008)3 studied the erties of apple trees with cytosporosis disease. Anjan Kumar
relations between climate, disease transmission and declines Payra et al. (2013)22 protein function for prediction.
in amphibians. Steinwart and A. Christmann et al. (2008)4
used the algorithm K -means to cluster the data set. Mruna-
Methodology
lini R. et al. (2011)5 identified clustering of K-means and
The basic steps of plant disease detection are shown in figure
artificial intelligence in pattern recognition of crop diseases.
1. The steps involved in our methodology are,
A. New spectral indices (NSIs) developed by K. Mahlein et
al. (2013)6. NSI is useful for the identification of crops of 1. Image Acquisition
various diseases. 2. Image Pre-processing
3. Image Segmentation
H. Cartwright et al. (2015)7 identified the Artificial Neural 4. Feature Extraction
Networks to simulate human brain model functioning. Raj 5. Classification
Kumar et al. (2016)8 addressed Hybrid Features and Auto- 1. Image Acquisition: The initial procedure is to collect
mated Leaf Disease Detection and Classification Supervised all infected or diseased plant leaf images via a camera
Classifier. They also built a database consisting of around or a source, which stores all infected plant leaf data.
32,000 different species of plants that are pre-processed, The first photo is in RGB type (Red, Green and Blue).
such as image reshaping, resizing and converting to a collec- In this colour transformation method, a device-inde-
tion. Related processing happens on the test image as well. pendent colour-space transition is generated for an
H. Azizpour et al. (2016)9 used Convolution networks to ex- RGB leaf image and then used for the colour transfor-
amine many factors affecting transferability, design, training mation method.
data distribution, feature extraction etc. 2. Image Pre-processing: The contaminated plant leaf
in the picture produces noise. The noise here may be
J. Garcia Arnal Barbedo et al. (2018)10 used the Annotated dust, on the leaf sand and other stuff. To get more ac-
Plant Pathology Databases for the identification and recog- curacy it is important to remove the noisy data from
nition of diseases based on photos. G. Zhou et al. (2019)11 the image. Pre-processing techniques are then used to
suggested, using FCM-KM and Faster R-CNN fusion, a eliminate noisy image images. There are several pre-
technique to resolve the various problems with rice disease processing techniques available some of which are
detection accuracy. Adnan Mushtaq et al. (2019)12 proposed Image clipping, i.e. cropping of the image of the leaf
a model for plant disease detection and offers solutions us- to get the area of interest. Smoothing of the image is

153 Int J Cur Res Rev | Vol 12 • Issue 20 • October 2020

 Pitchai et al.: Potato plant disease detection using convolution neural network

achieved with the smoothing filter. Improvement of for training the data. SSD_Mobilenet is a neural network
the picture is carried out to improve the contrast. The algorithm which is used for object detection. SSD is a
result it generates is dependent on any one of the tech- framework used for multi-box detectors and Mobilenet is
niques used. used for classification. Only by combining both, it is used
3. Image Segmentation: Segmentation means partition- for object detection. The job of mobile net layers is to con-
ing the image into separate parts with the same fea- vert the pixels from input image to feature that defines the
tures, or having some resemblance. The segmentation image and hence it is used as a feature extractor. With the
can be achieved using various approaches such as help of the mobile net, we can classify the image category.
clustering k-means, converting an RGB image to HIS The following is the procedure for training and testing the
model etc. Here, we used the algorithm K Means. The
model.
clustering of K-means is used to classify objects based
on a collection of features into the number of classes Initially, we collect a few hundred images that contain the re-
K. Object classification is achieved by minimizing the quired leaf images i.e., potato diseased leaf from an external
sum of the squares of the distance between the object source. We need to annotate the infected potato images by
and the cluster in the query. using a software called LabelImg because we need to draw
4. Feature Extraction: Extraction of features plays a a box around the infected region for the leaf image and must
key role in detecting a plant disease. Extraction func- specify the class for it to train it. This Label Img software
tion is used in many image processing applications. automatically will create an Extensible Markup Language
Colour, texture, shape, edges etc. are the character- (XML) file that describes the leaf image with infected region
istics that can be used in the identification of plant dimensions. We need to split the images data into train and
diseases. Morphological results are found to provide
test samples. The train samples are used for training the mod-
better results than the other features. Texture means
el and test samples are used for testing the model whether it
how the image is formed in colour, the roughness, the
is producing the required results or not.
image’s hardness. This method collects the character-
istics of tainted vegetable leaves. As we have discussed in the above statement LabelImg
5. Classification: After the extraction of the function is software creates an XML file but the model( SSD mobile
done, a classifier is created which is used to identify net) does not accept the XML file for training we need to
the plant disease. Depending on their leaves, the plant change the XML file to another file format so that we can
disease can be classified and is defined as healthy and use that file to train the model. We can convert the XML
unhealthy leaves. The unhealthy leaf is further graded file to comma-separated values (CSV) file by using some
based upon the type of disease with which it has been code. After the conversion of a file from XML to CSV use
attacked.
the generated train.csv file to train the model. The ssd_mo-
bilenet model uses tensor flow as a backend for object
RESULTS detection. It takes more time to train the model in normal
systems. So we have used google open-source cloud plat-
Our system is capable of reliably detecting diseases on po- form i.e., Google Colaboratory which provides GPU en-
tato plants. We need to deal with this with ample knowl- vironment and storage space up to an extent for the user.
edge of the disease and cure to improve the plant’s health. It is faster when compared to normal systems that do not
The framework implemented using python and the google have a Graphical Processor Unit (GPU) accessibility. The
GPU(Graphical Processor Unit). Figure 2 provides samples platform runtime is only 12 hours So we have to be careful
of the results of our system. Figure 2.a shows Healthy early while training the model. To train the predefined model we
blight potato leaf, Figure 2.b shows Healthy late blight po- need to change some code in the model configuration file
tato leaf and Healthy potato leaf Figure 2.c. Our system de- to our requirement and save it.
livered 80 per cent accuracy.
While training, the model generates some checkpoint files at
some interval of time so we need to save those checkpoints
DISCUSSION accordingly to start the training from that checkpoint even
The input for the model is photos of potato leaf that include if the platform runtime restarts. We can start the training
two disease types, one of which is early bright and the other from a particular checkpoint by using a checkpoint file by
late bright. The images input which is passed to the model is changing some code in it. In the ssd_mobilenet configuration
as follows. To get more accuracy for the model, the input im- file, there will be something like execution steps. We need
ages are preprocessed to eliminate noise data, dust, and other to mention a value by which the training ends whenever it
stuff from the images. reaches the specified execution step. The output of the model
is as follows in fig 2.
We applied the transfer learning technique to a predefined
convolutional neural network model i.e., SSD mobile net

Int J Cur Res Rev | Vol 12 • Issue 20 • October 2020 154
 Pitchai et al.: Potato plant disease detection using convolution neural network

CONCLUSION 5. Mrunalini R. Badnakhe. An Application of K-Means Clustering

and Artificial Intelligence in Pattern Recognition for Crop Dis-
Our proposed system was built taking into account the ben- eases, 2011 International Conference on Advancements in Infor-
efits of the farmers and the farming industry. The established mation Technology With workshop of ICBMG 2011: 134-138.
6. Mahlein AK et al. Development of spectral indices for detecting
system is capable of detecting disease in plants and also and identifying plant diseases, Remote Sensing of Environment.
providing the remedy that can be taken against it. We will 2011; 128: 21–30.
handle it by proper knowledge of the disease and the cure 7. Cartwright H, Ed., Artificial Neural Networks, Humana Press,
to improve the plant’s health. The framework proposed is 2015.
implemented using python and gives approximately 80 % 8. Raj Kumar S, Sowrirajan S. Automatic Leaf Disease Detection
and Classification using Hybrid Features and Supervised Classi-
accuracy. Use Google GPU to process will improve accuracy fier, Int. J. Adv. Res. Electrical, Electronics and Instrumentation
and speed. Shortly, we will be able to train the model with Engg. 2016; 5(6): 4556-4563.
more data that can cover all the diseases that are present and 9. Azizpour H et al. Factors of Transferability for a Generic Con-
it should be able to suggest the precautions. It can also be vNet Representation, IEEE Transactions on Pattern Analysis and
implemented in drones so that the drones capture the leaves Machine Intelligence, 2016; 38(9): 1790-1802.
10. Garcia J, Arnal B. Annotated Plant Pathology Databases for
from the top and detects the kind of disease that the plant has Image-Based Detection and Recognition of Diseases, in IEEE
affected and provide the necessary instructions for the farm- Latin America Transactions,2018; 16(6): 1749-1757.
ers to take further action to stop the disease. 11. Zhou G, Zhang W, Chen A, He M and Ma X. Rapid Detection
of Rice Disease Based on FCM-KM and Faster R-CNN Fusion,
IEEE Access, 2019;7: 143190-143206.
ACKNOWLEDGEMENT 12. Adnan Mushtaq et al. Plant Disease Detection using CNN &
Remedy, Int. J. Adv. Res. Electrical, Electronics and Instrumen-
We thank our colleagues from B.V. Raju Institute of Tech- tation Engg. 2010; 8(3): 622-626.
nology, Narsapur who provided insight and expertise that 13. Raghavendran S et al. Leaf Disease Classification using SVM
greatly assisted the research, although they may not agree Classifier in Cloud, Int. J. Innovative Techn. Exploring Engg.
(IJITEE), 2019; 9(1): 4146-4149.
with all of the interpretations/conclusions of this paper. We 14. Sankaran S, Mishra A, Ehsani R, and Davis C et al. A review
are also grateful to authors/ editors/publishers of all those of advanced techniques for detecting plant diseases, Computers
articles, journals and books from where the literature for this and Electronics in Agri. 2010; 72(1):1–13,
article has been reviewed and discussed. 15. Tyr Wiesner-Hanks et al. Millimeter-Level Plant Disease Detec-
tion From Aerial Photographs via Deep Learning and Crowd-
Conflict of Interest: There is no conflict of interest. sourced Data, Tech. Adv. Plant Sci. 2019; 10: 1-11.
16. Liu, L., Ouyang, W., Wang, X. et al. Deep Learning for Gener-
Financial Support: Not applicable ic Object Detection: A Survey. Int. J. Comput Vis, 2020; 128:
261–318.
Human or animal study Ethical clearance letter: Not Ap- 17. Pothen ME and Pai DM, Detection of Rice Leaf Diseases Us-
plicable ing Image Processing, 2020 Fourth International Conference
on Computing Methodologies and Communication (ICCMC),
Informed consent: Not Applicable
Erode, India, 2020; 424-430.
18. Sun J, Yang Y, He X and Wu X, Northern Maize Leaf Blight
Detection Under Complex Field Environment Based on Deep
REFERENCES Learning, in IEEE Access, 2020; 8: 33679-33688.
1. Van der Zwet, T. Present Worldwide Distribution Of Fire Blight. 19. Li J, C. Xu L, Jiang, Y, Xiao L, Deng Z and Han Z, Detection
Acta Hortic, 2002; 590: 33-34. and Analysis of Behavior Trajectory for Sea Cucumbers Based
2. Tatem, D. J. Rogers, and S. I. Hay et al. Global transport net- on Deep Learning, in IEEE Access, 2020; 8: 18832-18840.
works and infectious disease spread, Advances in Parasitology, 20. Chen W, Lin Y, Ng Y, Liu C and Lin Y. Rice Talk: Rice Blast
2006; 62: 293–343. Detection Using Internet of Things and Artificial Intelligence
3. Rohr JH, Raffel TR, Romansic JM, McCallum H, and P. J. Hud- Technologies, in IEEE Internet of Things Journal, 2020; 7:2:
son et al. Evaluating the links between climate, disease spread, 1001-1010.
and amphibian declines, Proceedings of the National Academy 21. Rasul Ruzmetov, Yulduz Matyakubova, Ikram Abdullaev. Cyto-
of Sciences of the United States of America. 2008; 105(45): sporosis Diseases of Apple Trees (Reinette Simirenkomalus) and
17436–17441. It’s Distribution in the Lower Amudarya Region International
4. Steinwart and Christmann R. Support Vector Machines, Spring- Journal of Current Research and Review. 2020;12 (14): 62-67.
er Science and Business Media, New York, NY, USA, 2008. 22. Payra, Anjan. Function Prediction using Cluster Analysis of
Unannotated Align Sequences. International Journal of Current
Research. 2013; 5: 134.

155 Int J Cur Res Rev | Vol 12 • Issue 20 • October 2020

 Pitchai et al.: Potato plant disease detection using convolution neural network

Figure 1: Basic steps for plant disease detection.

Figure 2: Sample results of disease detection.

Int J Cur Res Rev | Vol 12 • Issue 20 • October 2020 156