The objective of this project is to have a complete machine learning pipeline (training, validation, testing) based on the PyTorch Lightning framework.
PyTorch Lightning is presented to us as a PyTorch framework that allow us to right machine learning pipelines in a faster, cleaner and less complicated way without any cost in performance.
As a daily user of PyTorch, I wanted to experiment this framework to better understand its pros and cons. The best way to understand it is to get the hands dirty so I took a simple project based on PyTorch and I will "convert" it to PyTorch Ligthning.
Recently I found the project 2D Hand Pose Estimation from RGB Image by Olga Chernytska.
She explains the idea of the project in the following tutorials:
- Gentle introduction to 2D Hand Pose Estimation: Approach Explained
- Gentle introduction to 2D Hand Pose Estimation: Let's Code It!
The tutorials are written in a very detailed and clear way that I will not reproduce them here.
For this project, open-source FreiHAND dataset is used. The dataset is not part of this repository. It should be downloaded and moved to the root of the repo.
I was able to train and run a model using the Pytorch Lightning framework. This repo is closed.