paweszetela / image-classification

A CLI tool for rapid prototyping of PyTorch-based image classification models.

Image-Classification

A PyTorch-based project for image classification experiments, featuring rapid model prototyping through dynamic configuration.

Installation

Using Conda:

Create and activate the environment:

conda env create -f conda.yaml
conda activate image-classification

Using Pip:

Ensure a virtual environment (Python 3.9) is activated:

pip install -r requirements.txt

Usage

Configuration

This project uses Hydra for configuration managment. Modify main configuration file conf/config.yaml for custom experiment settings.

Running Experiments

From project root, execute the main script with optional configuration overrides:

python src/main.py training.n_epochs=5 experiment.use_gpu=False

Experiment Tracking with MLflow

Launch MLflow's UI for tracking experiments. Use optional --port for a custom port (default access at http://localhost:5000):

mlflow ui --port 8080