atar 1.0.15

A CLI tool for AI tasks such as dataset handling, training, and validation.

ATAR CLI

ATAR CLI is a command line interface that wraps common tasks for training and running YOLOv8 based fire detection models. It provides a simple interactive menu to download datasets, start and resume training, validate models and run live or video based detection.

Installation

The project can be installed from source using pip:

pip install -e .

The tool requires Python 3.7+ and depends on packages such as ultralytics, roboflow and supervision.

Usage

After installation run the CLI with:

atar-cli

The following menu options are available:

1. Download RoboFlow training dataset – Fetch a dataset from Roboflow.
2. Train – Start a new training run using a selected model and dataset.
3. Resume existing training – Continue a previous training run.
4. Validate – Evaluate a trained model on a validation dataset.
5. Live Test – Run detection on frames from a webcam.
6. Test on an existing file – Perform detection on a saved video file.
7. Quit – Exit the CLI.

Select an option by entering the corresponding number. Some options will ask for additional input such as dataset paths or model names.

Development

Install the development requirements and run the test suite with pytest to ensure everything works as expected.

pip install -r requirements.txt  # if available
pytest -q

Contributions are welcome! Feel free to open issues or pull requests on the GitHub repository.