Torchsight

WARNING: This package is currently under development.

Torchsight is a package that contains novel models for the few-shot object detection task.

This package was created to not only provide some novel models for object detection, it's also a framework to avoid boilerplate code that we usually see during the development of PyTorch experiments.

In this package you will find ready to use some Datasets, Loggers, Losses, Metrics, Optimizers, Trainers and Transforms.

If you are a pytorch user probably you already know what is a Dataset or an Optimizer, but the Logger and Trainer classes are new. This classes will help us to abstract the some tasks allowing us to reuse code in a beautiful way.

Quick start

TODO

Installation

The package is not ready yet, but if you want, you can clone the repo, create a conda environment with the dependencies and start using the models.

git clone https://github.com/SetaSouto/torchsight.git
cd torchsight
conda env create -f environment.yml

Then install the torchsight package in development mode:

python setup.py develop

Documentation

The documentation is autogenerated from the docstring using pdoc and you can visit it here.