CVR COLLEGE OF ENGINEERING
Department of Emerging Technologies
B.Tech CSE IV Year |I Semester
Technical Seminar,January,2024
Plant Disease Detection and Classification by
Deep Learning
Rishikesh Arra
21B81A0598
Under the guidance of
Dr M Varaprasad Rao
Abstract
• Deep learning is a key area of artificial intelligence focused on automatic learning and feature
ex tractio n.
•Widely used in image, video, voice, and natural language processing.
• Gaining attention in agriculture for plant disease recognition and pest assessment.
• Offers objective disease detection, improving efficiency and technology adoption.
• The review highlights recent progress, trends, and challenges in using deep learning for plant disease
detection.
•Aims to be a valuable resource for agricultural researchers while addressing current challenges.
Introduction
• Early detection of plant diseases is essential for effective management and to prevent food insecurity.
• Traditional methods, where experts or farmers visually identify diseases, are often subjective,
inefficient, and time-consuming.
• Inexperienced farmers may misjudge diseases, leading to the incorrect application of chemicals, which
can result in environmental pollution and unnecessary economic losses.
• There is a growing need for automated, accurate, and efficient methods to detect plant diseases and
reduce human error in agriculture.
• Develop a robust, automated system for plant disease detection that reduces the subjectivity and
inefficiencies associated with traditional visual identification methods used by farmers and agricultural
ex p erts .
•
Apply image processing techniques like K-means clustering to identify diseased plant areas.
•
Use machine learning methods, such as improved Support Vector Machines (SVM), to classify and
detect plant diseases with high accuracy (up to 93%).
Existing Work
•Chai et al. achieved 94.71% and 98.32% accuracy for tomato leaf diseases using discriminant analysis,
while Li and He used a BP neural network for apple leaf diseases with 92.6% accuracy. Guan et al.
reached 97.2% accuracy for rice diseases using Bayesian methods.
Drawbacks
• Time-Consuming: Traditional methods rely on manual feature extraction, making them labor-intensive.
• Spot Segmentation Dependence: Heavy reliance on accurate disease spot identification reduces
r eliab ility.
• Limited Generalizability: Models perform poorly in complex environments.
Goals
• Achieve over 90% accuracy in detection.
• Ensure the system works in real-world agricultural settings.
Outcomes
• Faster, more accurate disease detection.
• Reduced environmental and economic impact through better diagnosis.
Literature Review
Chai et al.:
• Studied four tomato leaf diseases and extracted 18 characteristics (color, texture, shape).
• Used PCA and Fisher discriminant analysis, achieving 94.71% and 98.32% accuracy.
Li and He:
• Investigated 5 apple leaf diseases using 8 features (color, texture, shape).
• Applied BP neural network, achieving 92.6% average accuracy.
Guan et al.:
• Analyzed 63 parameters of rice diseases (morphology, color, texture).
• Used discriminant analysis, achieving 97.2% accuracy.
Architecture diagram
Proposed method
Dataset and Augmentation:
• Use datasets like PlantVillage, applying data augmentation and GANs to expand small samples.
Deep Learning Model:
• Employ CNNs (e.g., ResNet, Inception) for disease classification, with Few-Shot Learning or
Transfer Learning for small datasets.
Advanced Imaging:
• Integrate Hyperspectral Imaging (HSI) and 3D-CNNs to enhance early disease detection.
Visualization:
• Use saliency maps or Grad-CAM to interpret model predictions.
Evaluation:
•Assess performance using precision, recall, F1-score, accuracy, and mAP for multi-disease scenarios.
RESULTS AND DISCUSSION
•GANs improved disease detection accuracy by enhancing dataset size,
e.g., AR-GAN boosted accuracy by 5.2%.
•CNN Models like GoogLeNet achieved high accuracy (94.33%) for
plant disease classification.
•Hyperspectral Imaging significantly improved early detection,
reaching 97.1% accuracy in some cases.
•Transfer Learning with models like VGG16 and ResNet50 enhanced
accuracy on small datasets.
References
[1] F. Fina, P. Birch, R. Young, J. Obu, B. Faithpraise, and C. Chatwin,‘Automatic plant pest detection and recognition
using k-means clustering algorithm and correspondence filters,’’ Int. J. Adv. Biotechnol. Res., vol. 4, no. 2, pp. 189–
199, Jul. 2013.
[2] M. A. Ebrahimi, M. H. Khoshtaghaza, S. Minaei, and B. Jamshidi,‘‘Vision-based pest detection based on SVM
classification method,’’Comput. Electron. Agricult., vol. 137, pp. 52–58, May 2017.
[3] S. R. Dubey and A. S. Jalal, ‘‘Adapted approach for fruit disease identi-fication using images,’’ Int. J. Comput. Vis.
Image Process., vol. 2, no. 3,pp. 44–58, Jul. 2012.
[4] A.-L. Chai, B.-J. Li, Y.-X. Shi, Z.-X. Cen, H.-Y. Huang, and J. Liu,‘‘Recognition of tomato foliage disease based on
computer vision tech_x0002_nology,’’ Acta Horticulturae Sinica, vol. 37, no. 9, pp. 1423–1430,Sep. 2010.
[5] Z. R. Li and D. J. He, ‘‘Research on identify technologies of apple’sdisease based on mobile photograph image
analysis,’’ Comput. Eng. Des.,vol. 31, no. 13, pp. 3051–3053 and 3095, Jul. 2010.
Thank you
