pytorch-polkadots 0.1.0

Provides the ability to construct a simulated image dataset on the fly with different characteristics.

Overview

This project provides the ability to construct a simulated image dataset on the fly with different characteristics. While it is possible to simulate datasets that models can be trained against successfully, it is also useful to adjust dataset parameters to yields datasets that much more challenging to train against.

Installation

With poetry

poetry install 

With pip

pip install pytorch-polkadots 

From cloned repo

pip install -r requirements.txt

Tests

bash
export PYTHONTEST=.
pytest

Example image

You can see an image generated by the code with:

python main.py

This will display the following image:

PyTorch dataset

The code provides a pytorch dataset that you can use in benchmarks.

class PolkaDotDataset(Dataset)