Computer Vision Talks
Posts Tags Open-Source Competitions About Research
Computer Vision Talks

Open-Source

/images/github_activity_2020.png

You can find full list of my open-source projects under my GitHub account: https://github.com/BloodAxe. However, there are some projects I worked on that are worth mentioning here:

PyTorch-Toolbelt

https://github.com/BloodAxe/pytorch-toolbelt

Pytorch-toolbelt is a Python library with a set of bells and whistles for PyTorch for fast R&D prototyping and Kaggle farming.

What’s inside:

  • Easy model building using flexible encoder-decoder architecture.
  • Modules: CoordConv, SCSE, Hypercolumn, Depthwise separable convolution and more.
  • GPU-friendly test-time augmentation TTA for segmentation and classification
  • GPU-friendly inference on huge (5000x5000) images
  • Every-day common routines (fix/restore random seed, filesystem utils, metrics)
  • Losses: BinaryFocalLoss, Focal, ReducedFocal, Lovasz, Jaccard and Dice losses, Wing Loss and more.
  • Extras for Catalyst library (Visualization of batch predictions, additional metrics)
  • Showcase: Catalyst, Albumentations, Pytorch Toolbelt example: Semantic Segmentation @ CamVid

Why Honest answer is “I needed a convenient way to re-use code for my Kaggle career”. During 2018 I achieved a Kaggle Master badge and this been a long path. Very often I found myself re-using most of the old pipelines over and over again. At some point it crystallized into this repository.

This lib is not meant to replace catalyst / ignite / fast.ai high-level frameworks. Instead it’s designed to complement them.

Albumentations

https://github.com/BloodAxe/albumentations

This project is a fork of Albumentations. I was a member of Albumentations core team back in a day. Since the beginning of russian invasion into Ukraine I’ve left the team. You can find answers why here.

This fork contains few improvements/additions that I feel are necessary to have. Since this private fork is mainly meant for personal use I feel less obliged to keep backward compatibility with the original project.

At some point it will become deprecated and transform into something new & shiny.

Catalyst

https://github.com/bloodAxe/catalyst

Yet another fork of high-level training library around PyTorch. Forked quite a while ago to ensure API will remain stable and also added missing features to my taste.