Title: Plant Disease Detection

Technologies Used :

Data Analysis Using Python,
Computer Vision,
Deep Learning (CNN)

Commands For Installing Libraries :

pip install numpy pandas opencv-python
pip install torch torchvision torchaudio
pip install pillow
pip install streamlit

Commands For Running Project :

python main.py
python app.py
streamlit run app.py

To Clone :

git clone https://github.com/shangarwarsandesh/Plant-Disease-Detection.git

Dataset Links :

'Dataset Reference Link'

Steps to Develop Project :

1. Define the Objective

• Clearly outline the goals of the project, e.g., identifying specific plant diseases from leaf images or detecting general unhealthy conditions in plants.

2. Data Collection

• Gather a dataset of plant images, including healthy anddiseased plants. Sources: Online repositories (e.g., Kaggle, PlantVillage), field data, or smartphone images.
• Ensure diverse images with varying lighting, angles, and environments for robustness.

3. Data Preprocessing

• Label the Data: Annotate images with disease categories or "healthy" labels.
• Clean the Data: Remove duplicates, irrelevant, or low-quality images.
• Augment the Data: Apply transformations like flipping, rotation, scaling, and color variations to increase dataset size and diversity.
• Resize Images: Standardize the dimensions to reduce computational requirements.

4. Model Selection

Choose a suitable computer vision approach:

• Pretrained Models: Use models like ResNet-18.

5. Model Training

• Split the dataset into training, validation, and test sets (e.g., 70%-20%-10%).
• Train the model using the training set, tuning hyperparameters like learning rate, batch size, and epochs.
• Validate the model periodically to check for overfitting or underfitting.
• Use techniques like early stopping and dropout to improve performance.

6. Model Evaluation

Test the model on the test dataset to evaluate metrics such as:

• Accuracy: Overall prediction correctness.
• Precision/Recall/F1 Score: For imbalanced datasets.
• Confusion Matrix: For detailed performance on each class.

7. Deployment Preparation

Optimize the model for real-world usage:

• Reduce size using quantization or pruning.
• Convert to formats suitable for deployment (e.g., TensorFlow Lite, ONNX).

8. Build a User Interface

Create a simple interface for end-users using Streamlit.

9. Testing and Validation

• Test the system in real-world scenarios to identify potential issues.
• Gather feedback from end-users to improve usability and accuracy.

10. Deployment and Maintenance

• Deploy the system on cloud servers, mobile apps, or local devices.
• Monitor its performance, update the model with new data, and fix bugs as needed.

11. Documentation and Reporting

• Document the project comprehensively, including methods, results, and challenges.
• Prepare a report or presentation for stakeholders.

12. Future Enhancements

• Incorporate more features, such as real-time video nalysis or multi-plant disease detection.
• Continuously update the dataset and retrain the model to improve accuracy.

Authors :

Sandesh Shangarwar