🥔 Potato Disease Classification with CNN

This project uses a Convolutional Neural Network (CNN) to classify potato leaf images into three categories:

• Potato___Early_blight
• Potato___Late_blight
• Potato___healthy

The model is trained using TensorFlow and Keras on the PlantVillage dataset.

---

📂 Dataset

• Total Images: 2152
• Classes: Early Blight, Late Blight, Healthy
• Image Size: 256x256
• Splits:
  • Training: 80%
  • Validation: 10%
  • Testing: 10%

---

🧠 Model Architecture

Sequential([
    Resizing(256, 256),
    Rescaling(1./255),
    Conv2D(32, (3,3), activation='relu'),
    MaxPooling2D(),
    Conv2D(64, (3,3), activation='relu'),
    MaxPooling2D(),
    Flatten(),
    Dense(128, activation='relu'),
    Dense(3)  # 3 output classes
])

• Optimizer: Adam
• Loss: Sparse Categorical Crossentropy
• Metric: Accuracy
• EarlyStopping: Enabled to prevent overfitting

---

📊 Evaluation

• Accuracy and loss are tracked for both training and validation sets.
• A sample batch of predictions (9 images) is visualized along with their predicted and actual labels.

---

🚀 How to Run

1. Clone the repository and install dependencies:

   pip install -r requirements.txt

2. Ensure the dataset folder path is:

   ../PlantVillagePotatoDataset/
   

3. Run the Jupyter notebook:

   jupyter notebook model.ipynb

---

📝 Output Sample

Displays predictions with confidence scores and actual labels for 9 sample test images.

---

📌 .gitignore Suggestions

• Ignore model files, datasets, cache, logs, and system files to keep the repository clean.

---

🌐 Web App: Streamlit Interface

A user-friendly web application is provided for potato disease classification using Streamlit.

Features

• Upload a potato leaf image (JPG, JPEG, PNG)
• The app displays the uploaded image and predicts its condition: Healthy, Early Blight, or Late Blight
• Shows the prediction confidence score
• Clean, centered UI with clear result display

How to Run the Web App

1. Make sure you have all dependencies installed (see requirements.txt)
2. Ensure the trained model file is available at: ../models/potato_disease_classification_v1.0.h5
3. From the server/ directory, run:

   streamlit run server.py

4. Open the provided local URL in your browser to use the app.

App UI Overview

• The main page allows you to upload a potato leaf image.
• The left panel shows the input image.
• The right panel displays the predicted disease class and confidence.
• Supported classes: Early Blight, Late Blight, Healthy

---

🔮 Future Development

• XAI Integration:
  • Plan to integrate Explainable AI (XAI) techniques (such as Grad-CAM ) to provide visual explanations for model predictions, helping users and researchers understand which parts of the leaf image influenced the classification decision.