Skip to content

INSTALLATION.md

Latest commit

 

History

History
135 lines (101 loc) · 3.3 KB

INSTALLATION.md

File metadata and controls

135 lines (101 loc) · 3.3 KB

Garbage Classification - Installation Guide

Prerequisites

  • Python 3.8+ (3.9-3.11 recommended)
  • 8GB+ RAM (16GB recommended for training)
  • 5GB+ storage for datasets and models
  • Git and pip installed

Verify installation:

python --version  # Should show 3.8+
pip --version

Quick Setup

# 1. Fork repository (if contributing)
# Go to https://github.com/AditixAnand/Garbage_Classification.git
# Click "Fork" button to create your own copy

# 2. Clone repository (use your fork URL if contributing)
git clone https://github.com/YOUR_USERNAME/Garbage_Classification.git
# Or clone original:
# git clone https://github.com/AditixAnand/Garbage_Classification.git
cd Garbage_Classification

# 3. Create virtual environment
python -m venv .venv

# 4. Activate environment
# macOS/Linux:
source .venv/bin/activate
# Windows:
.venv\Scripts\activate

# 5. Install dependencies
pip install -r requirements.txt

# 6. Start Jupyter
jupyter notebook

Dataset Setup

Automatic (Recommended): Run the KaggleHub download cell in any notebook - dataset downloads automatically to ~/.cache/kagglehub/

Manual: If automatic fails, organize your dataset as:

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

Running Notebooks

Execution Order:

  1. Week_1.ipynb - Data exploration and basic preprocessing
  2. Week_2.ipynb - EfficientNetV2B2 model training + Gradio interface
  3. Week_3.ipynb - Advanced training and deployment

Alternative to Jupyter: Use VS Code with Python extension for notebook editing.

Key Dependencies

tensorflow>=2.13.0      # Deep learning framework
gradio==4.44.0          # Web interface
scikit-learn>=1.3.0     # ML utilities
matplotlib>=3.7.0       # Visualization
kagglehub>=0.2.0        # Dataset downloads

Model Architecture

  • Base: EfficientNetV2B2 with ImageNet weights
  • Input: 124×124×3 RGB images
  • Classes: 6 (cardboard, glass, metal, paper, plastic, trash)
  • Target Accuracy: >95%
  • Deployment: Gradio web interface

Troubleshooting

Common Issues:

# TensorFlow GPU (M1/M2 Mac)
pip install tensorflow-macos tensorflow-metal

# Memory errors - reduce batch size in notebooks
batch_size = 16  # instead of 32

# Package conflicts
pip uninstall -r requirements.txt -y
pip install -r requirements.txt

# Virtual environment issues (Windows)
.venv\Scripts\activate.bat

Port Conflicts: Notebooks auto-find available ports. Restart kernel if needed.

Project Structure

Garbage_Classification/
├── Week_1.ipynb         # Data exploration
├── Week_2.ipynb         # EfficientNet training
├── Week_3.ipynb         # Advanced deployment
├── requirements.txt     # Dependencies
├── INSTALL.md          # This guide
└── .venv/              # Virtual environment

Development

Environment Setup:

# Always activate before working
source .venv/bin/activate  # macOS/Linux
.venv\Scripts\activate     # Windows

# Verify TensorFlow
python -c "import tensorflow as tf; print(tf.__version__)"

Contributing: Fork → Create branch → Follow setup → Make changes → Submit PR

Need Help? Create an issue on GitHub with your error details and system info.