Trash Classification

Project Overview

This project aims to classify different types of trash using a deep learning model. The model is trained on the garythung/trashnet dataset and uses a custom architecture called Trashmobilenet-v1.

Project Structure

.
├── checkpoints
│   └── best_model.pth
├── LICENSE
├── main.py
├── Makefile
├── notebook
│   └── experiment_trash_classification.ipynb
├── pyproject.toml
├── README.md
├── requirements.txt
├── src
│   ├── callback
│   │   └── earlystopping.py
│   ├── config.py
│   ├── data
│   │   └── preprocessing.py
│   ├── metric
│   │   └── metric.py
│   └── models
│       ├── evaluation.py
│       ├── model.py
│       └── training.py
├── test.py
└── uv.lock

Requirements

• Python 3.12 or higher
• See requirements.txt for a full list of dependencies

Setup

1. Clone the repository:

   git clone [email protected]:pradanaadn/trash-detection.git
   cd trash-detection

2. Create and activate a virtual environment:

   python3 -m venv .venv
   source .venv/bin/activate

3. Install the dependencies:

   pip install -r requirements.txt

4. Login to Weights & Biases:

   wandb login <your_wandb_api_key>

5. Login to Hugging Face:

   huggingface-cli login --token <your_hf_token>

Running the Project

To train the model, run:

python main.py