Detection of Tomato Leaf Disease

Locations Using Deep Learning

PURPOSED BY :-
RAVI KUMAR SHAH
RA2012005010027
CSE MTECH
ABSTRACT

⮚To improve the recognition model accuracy of crop disease leaves and locating
diseased leaves, this paper proposes an improved Faster RCNN to detect healthy
tomato leaves.
⮚ First, we use a depth residual network to replace VGG16 for image feature
extraction so we can obtain deeper disease features.
⮚Second, the k-means clustering algorithm is used to cluster the bounding boxes.
We improve the anchoring according to the clustering results. The improved
anchor frame tends toward the real bounding box of the dataset.
⮚Finally, we carry out a k-means experiment with three kinds of different feature
extraction networks.
⮚The experimental results show that the improved method for crop leaf disease
detection had 2.71% higher recognition accuracy and a faster detection speed
than the original Faster RCNN.(previous it has 97.31 and In Faster RCNN-res101,
the mAP after using k-means analysis increased by 1.36% to 98.54%)
Research problem statement
⮚Crop disease detection is the basis of crop disease prevention to guarantee crop
quality.
⮚Traditional detection methods for crop disease mainly depend on manual
observation and consequently lead to low detection efficiency and poor
reliability.
⮚Farmers lack professional knowledge, and agricultural experts cannot serve the
field at all times so that they miss the best time for prevention.
⮚In recent years, image processing , pattern recognition , computer vision and
other technologies have developed rapidly. Computer automatic detection of
diseases provides method for effectively solving agricultural problems.
LITERATURE SURVEY :
Title of the Paper / Author/ Pros Cons
Year
of Publication
Tomato Leaf Disease Detection Using Deep This paper proposes to identify the Tomato Plant The proposed work can be further extended for
Learning Techniques Leaf disease using image processing techniques various new algorithms to provide optimum
Proceedings of the Fifth International based on Image segmentation, clustering, and results in context to existing techniques. Real-
Conference on Communication and Electronics open-source algorithms, thus all contributing to a time application based plant disease
Systems reliable, safe, and accurate system of leaf categorization will be one of the main factors in
(ICCES 2020) disease with the specialization to Tomato Plants. the selection of the appropriate technique.
The proposed method is achieved an accuracy
level of 97%.

Tomato Leaf Disease Diagnosis and Severity this paper we have proposed a segmentation- In future, we need try to improve the results in
Measurement based approach for automatic segmentation of terms of accuracy by combining different
Haridas D. Gadade Government College of infected regions. The segmented area is further features. Another way of the future would be to
Engineering, Jalgaon,Maharashtra-India analyzed for disease classification and severity provide with more robust segmentation method.
(2018) measurement. Leaf disease detection technique This method can also be extended to other
proposed here involves various stages including crops.
preprocessing, segmentation, feature extraction,
training and classification followed by the
severity measurement from the disease
segmented region.
Title of the Paper / Author/ Pros Cons
Year
of Publication
Identification of Plant Disease using Image It involves loading an image, image we might prefer to extend our work additional on a
Processing Technique. preprocessing, image segmentation, feature lot of disease detection
International Conference on Communication and extraction and classification. Development of
Signal Processing, April 4-6, 2019, India automatic detection system using advanced
technology like image process facilitate to support
the farmers within the identification of diseases at
an early or initial stage and supply helpful data for
its management.
An application of image processing techniques for The study of interest is the leaf rather than whole The proposed work can be further extended for
detection of diseases on brinjal leaves using k-means brinjal plant because about 85-95 % of diseases various new algorithms to provide optimum results
clustering method. occurred on the brinjal leaf like, Bacterial Wilt, in context to existing techniques. Real-time
Cercospora Leaf Spot, Tobacco mosaic virus (TMV). The
2019 methodology to detect brinjal leaf disease in this work application based plant disease categorization will
includes K-means clustering algorithm for be one of the main factors in the selection of the
segmentation and Neural-network for classification. appropriate technique.
The proposed detection model based artiifical neural
networks are very effective in recognizing leaf
diseases.
PURPOSED WORK
⮚The two improved methods are combined to detect tomato diseases. The
experiment shows that our approach can both improve the accuracy of tomato
disease recognition and detect the position of diseased leaves.
⮚This paper makes the following contributions:
• We use the residual network instead of VGG16 to extract features in Faster
. RCNN .ResNet can extract deeper tomato disease features than VGG16.
• We use the k-means algorithm to cluster the bounding boxes of datasets and
improve the anchor in Fast-RCNN according to the clustering results.
⮚Our method improves the feature extraction network and anchor. In feature
extraction, we use resnet to extract features. It can both extract the deep disease
characteristics of tomato disease and solve the problem of gradient
disappearance caused by the increase of layers.
⮚In region proposal network (RPN), we use k-means to cluster the bounding boxes
of tomato disease datasets. We improve anchor by clustering results.
CONCLUSION
⮚This paper improves the Faster RCNN algorithm to detect diseased tomato leaves,
which can both recognize tomato diseases and detect tomato leaf locations.
⮚To make the anchors in the algorithm closer to the ground truth of our dataset,
we use the k-means algorithm to cluster the bounding boxes of tomato disease
images and improve the anchors based on the results.
⮚We selected ResNet101 to replace VGG16 for feature extraction, which can
extract the deep features of tomato disease.
⮚The experimental results show that our method can effectively detect and
recognize tomato diseases and has higher detection accuracy than the original
Faster RCNN. However, the dataset is laboratory data, so only a single leaf disease
in the image can be detected.
⮚Future works will include images of natural plants that should be collected for
detection. In addition, tomato diseases have other disease characteristics, such as
fruits and stems. Some diseases are a combination of these characteristics.
⮚Future works should combine these factors to carry out multifaceted diagnoses,
which will help the development of smart agriculture.
questionnaire
THANK YOU