Atayaz/AugmentedGrapevineDiseaseCNN

🍇 Grape Disease Classification with CNN and Transfer Learning

This project implements two deep learning models to classify grape leaf diseases:

1. 1.`grapevine_cnn.py` – A Convolutional Neural Network (CNN) built from scratch
2. 2.`grape_transfer_resnet.py` – A Transfer Learning model using ResNet50

Both models are deployed using Streamlit, allowing for training visualization and image-based prediction.

📁 Dataset

Dataset Used: Augmented Grape Disease Dataset Folder Name: Final Training Data

Each subfolder contains images labeled for a specific grape disease category:

• —Black Rot
• —ESCA
• —Leaf Blight
• —Healthy

⚙️ How to Run

Install required packages:

pip install -r requirements.txt

Run CNN model:

streamlit run grapevine_cnn.py

Run Transfer Learning model:

streamlit run grape_transfer_resnet.py

🧠 Model 1: CNN (grapevine_cnn.py)

Key Features:

• —5 Convolutional layers + 3 MaxPooling layers
• —Dropout layers for regularization
• —Precision, Recall, and Accuracy metrics
• —Uses train_test_split for dataset split
• —Accuracy and Loss visualizations with Matplotlib
• —Streamlit interface with image upload for prediction
• —Model saved as grape_cnn_model.h5

🧠 Model 2: Transfer Learning (grape_transfer_resnet.py)

Key Features:

• —Based on pre-trained ResNet50 (ImageNet weights)
• —Frozen convolutional base layers
• —Classification head with GlobalAveragePooling2D + Dense layers
• —Uses ImageDataGenerator with augmentation and validation split
• —EarlyStopping to prevent overfitting
• —Accuracy and Loss training curves
• —Streamlit image upload and prediction interface
• —Model saved as grape_resnet_model.h5

🖼️ Streamlit Interface Includes:

• —Real-time training graphs: Accuracy & Loss
• —Upload your own grape leaf image for prediction
• —Displays predicted disease class with confidence

💾 Model Output Files