LeafDiagnosis: Plant disease Detection.pptx

Project on plant disease detection

1. LeafDiagnosis: Deep Learning from Scratch JAHNAVI M.Sc. Int. (CS) 2303862

2. Contents Introduction Problem Definition Overview Of Proposed System Literature Survey Software Requirements P roposed Methodology About Dataset Objectives Conclusion References

3. Introduction In India,70% of population depend on agriculture and contributes 17% towards the GDP. Farmers experience great difficulties in switching from one disease control policy to another. The naked eye observation of experts is the traditional approach, this method can be time consuming, expensive and inaccurate. The crop losses can be minimized by applying pesticides or its equivalent to combat the effect of specific pathogens, if diseases are correctly diagnosed and identified early.

4. 1.PROBLEM DEFINITION: Agriculture crops are threatened by wide variety of plant diseases. These can damage the crop, lower the vegetable and fruits quality and wipe out the harvest. About 42% of the world's total agricultural crop is destroyed yearly by diseases.

5. 2.Overview Of Proposed System Innovation : Convolutional Neural Network Benefit : Increase in accuracy rate Classification of various plant diseases Advancement : Pesticide suggestion

6. 3.Literature Survey S.N o. PAPER TITTLE & PUBLICA T ION DETAILS NAME OF THE AUTHORS TECHNICAL IDEAS / ALGORITHMS USED IN THE PAPER & ADVANTAGES SHORTFALLS/DI S AD VANTAGES & SOLUTION PROVIDED BY THE PROPOSED SYSTEM 1. W. M. Fadzil, S. Rizam, J. R., and N. M. T., “Orchid leaf disease detection using Border Segmentation technique,” in Proc. IEEE Conf. Systems, Process and Control (ICSPC), vol. 1, pp. 168-179, Dec. 2014. 1.W.M.N Wan Mohd Fadzil 2.M.S.B. Shah Rizam 3.R. Jailani 4.M. T. Nooritawati Image segmentation technique Border segmentation techniques in MATLAB Adv:Accurate and automated classification of black leaf spot and sun scorch in orchid leaves using MATLAB with high accuracy. Limited to only two types of orchid leaf diseases and moderate accuracy. Solutions: Combining multiple preprocessing and morphological techniques to improve segmentation accuracy.

7. Literature Survey S.N o. PAPER TITTLE & PUBLICA T ION DETAILS NAME OF THE AUTHORS TECHNICAL IDEAS / ALGORITHMS USED IN THE PAPER & ADVANTAGES SHORTFALLS/DISAD VANTAGES & SOLUTION PROVIDED BY THE PROPOSED SYSTEM 2. V. Jumb, M. Sohani, and A. Shrivas, “Color image segmentation using K-means clustering and Otsu’s adaptive thresholding,” Int. J. Innovative Technol. Exploring Eng., vol. 3, no. 9, pp. 33-39, Feb. 2014. Vijay Jumb, Mandar Sohani, Avinash Shrivas Color Space Conversion, Value Channel Extraction, Otsu’s Multi-thresholding, K-means Clustering, Morphological Processing, PSNR and MSE Metrics. Adv:Improved segmentation accuracy and effective for various images. Over-segmentation and increased computational complexity. Solutions:Integration with K-means clustering reduces over-segmentation; morphological processing refines segmentation.

8. Literature Survey S.No. PAPER TITTLE & PUBLICA T ION DETAILS NAME OF THE AUTHOR S TECHNICAL IDEAS / ALGORITHMS USED IN THE PAPER & ADVANTAGES SHORTFALLS/DISAD VANTAGES & SOLUTION PROVIDED BY THE PROPOSED SYSTEM 3. F. Marzougui, M. Elleuch, and M. Kherallah, "A Deep CNN Approach for Plant Disease Detection," in 2020 21st International Arab Conference on Information Technology (ACIT), pp. 1-6, 2020. 1.Fatma Marzougui 2.Mohamed Elleuch 3.Monji kherallah Uses the ResNet and data argumentation on data , ResNet has more accuracy and less training time than CNN Without set background the accuracy is less

9. Literature Survey S.N o. PAPER TITTLE & PUBLICA T ION DETAILS NAME OF THE AUTHOR S TECHNICAL IDEAS / ALGORITHMS USED IN THE PAPER & ADVANTAGES SHORTFALLS/DISAD VANTAGES & SOLUTION PROVIDED BY THE PROPOSED SYSTEM 4. A. KP and J. Anitha, "Plant disease classification using deep learning," 2021 3rd International Conference on Signal Processing and Communication (ICPSC), 2021, pp. 407-411, doi: 10.1109/ICSPC51351.2021.9451696. 1.Akshai KP 2. J. Anitha The literature study reveals that pre-trained models using transfer learning are an efficient strategy for plant disease classification The ResNet model adds the output of one layer to the next layer that’s why the accuracy is less than the densenet .

10. 4.Software Requirements Operating System : Windows 10 IDE : python 3.7, MATLab , TensorFlow Language : Python

11. 5. PROPOSED METHODOLOGY The main steps in the proposed work is summarized below: • This work proposes a training image generation technology based on image processing techniques, which can enhance the robustness and prevent overfitting of the CNN-based model in the training process. • A convolutional neural network is first employed to diagnose leaf diseases; the end-to -end learning model can automatically discover the discriminative features of the leaf images and identify the common types of leaf diseases with high accuracy.

12. Methodology Problem Definition and Objective Setting: Define the scope and goals of the project, focusing on early detection of plant diseases to facilitate better treatment and crop management. Data Collection: Collect images of healthy and diseased plant leaves from various sources, including field data collection, online repositories, and research databases. Ensure a diverse and comprehensive dataset, capturing different types of diseases and varying stages of disease progression.

13. Methodology 3. Image Preprocessing: Loading Images: Use libraries like OpenCV or PIL to load images into a consistent format. Resizing and Normalizing: Resize images to a consistent size and normalize pixel values to thespecific range. Data Augmentation: Apply techniques such as rotations, flips, zooms, and brightness adjustments to increase dataset variability and improve model robustness. Noise Reduction and Contrast Enhancement: Use techniques like Gaussian blur and histogram equalization to improve image quality. 4. Model Selection and Preparation: Custom Model : Design and implement a custom convolutional neural network (CNN) from scratch for comparison.

14. Methodology 5. Model Training: Custom Model: Training the custom CNN using the preprocessed dataset, experimenting with different architectures, hyperparameters, and optimization techniques. 6. Model Evaluation: Analyzing confusion matrices to understand model performance in detail, identifying strengths and weaknesses in disease detection. 7. Deployment: Develop a user-friendly interface (e.g., web or mobile application) to allow users to upload leaf images and receive disease diagnoses. Implement the deployed model using frameworks such as Flask or Django for web applications, ensuring scalability and accessibility.

15. • Convolutional Neural Network (CNN) A convolutional neural network (CNN or ConvNet) is a network architecture for deep learning that learns directly from data. Purpose: CNNs are used to analyze visual data like images and videos. Convolutional Layers: These layers scan the input image with filters to detect patterns like edges, colors, or textures. Activation Functions: Functions like ReLU are applied after convolution to add non-linearity, helping the network learn complex patterns. Pooling Layers: These layers reduce the size of the data, making the network faster and more efficient while also making it better at recognizing patterns regardless of their location. Fully Connected Layers: At the end, these layers combine all the learned features to make a final decision or prediction. Dropout Layers: These layers help prevent overfitting by randomly ignoring some neurons during training, making the network more robust.

16. 6. About Dataset This dataset consists of about 87K rgb images of healthy and diseased crop leaves which is categorized into 38 different classes. The total dataset is divided into 80/20 ratio of training and validation set preserving the directory structure. A new directory containing 33 test images is created later for prediction purpose

17. 7. Objectives: The model automates and improves the classification of plant diseases from images using convolutional neural networks (CNNs). It achieves accurate and reliable predictions for effective disease management. Facilitates early detection of plant diseases, leading to timely and effective treatment. Develops a user-friendly platform for monitoring plant health. Contributes to research by providing publicly available trained models for further advancements. Aims to enhance crop health and reduce economic losses.

18. 6.Conclusion LeafDiagnosis: Deep Learning from Scratch helps in diagnosing the disease of the plants by scanning the leaves of the respective plant to help the agricultural field. It is based upon the Deep Learning technique and by leveraging Convolutional Neural Networks (CNNs), the model provides an effective and automated approach to classify plant diseases from images of leaves. It is a fast, accurate, and cost-effective method for disease detection.

19. 7. References [1] W. M. Fadzil, S. Rizam, J. R., and N. M. T., “Orchid leaf disease detection using Border Segmentation technique,” in Proc. IEEE Conf. Systems, Process and Control (ICSPC), vol. 1, pp. 168-179, Dec. 2014. [2] V. Jumb, M. Sohani, and A. Shrivas, “Color image segmentation using K-means clustering and Otsu’s adaptive thresholding,” Int. J. Innovative Technol. Exploring Eng., vol. 3, no. 9, pp. 33-39, Feb. 2014. [3] F. Marzougui, M. Elleuch, and M. Kherallah, "A Deep CNN Approach for Plant Disease Detection," in 2020 21st International Arab Conference on Information Technology (ACIT), pp. 1-6, 2020. [4] A. KP and J. Anitha, "Plant disease classification using deep learning," in 2021 3rd International Conference on Signal Processing and Communication (ICPSC), pp. 407-411, 2021.

20. THANK YOU