Garbage Classification using PyTorch

Stepwise Project Structure

Step 1 – Install Dependencies

Use the following command in a fresh Python 3.11 environment:

pip install -r requirements.txt

Step 2 – Technologies Used

• Python
• PyTorch
• Torchvision
• Scikit-learn
• Pillow
• Matplotlib
• tqdm

Step 3 – Dataset Format

Organize dataset in this structure:

data/
├── cardboard/
├── glass/
├── metal/
├── paper/
├── plastic/
└── trash/

Step 4 – Model

• Custom CNN (Convolutional Neural Network) neural network built from scratch
• GPU accelerated training using CUDA

Step 5 – Output

• Saved trained model: garbage_model.pth
• Performance metrics: accuracy, precision, recall, f1-score

Step 6 – Future Scope

• Improve dataset size and balance
• Use deeper CNN or ensemble techniques to target 96–99% accuracy

---