search

Westlake-AI / openmixup

CAIRI Supervised, Semi- and Self-Supervised Visual Representation Learning Toolbox and Benchmark
https://openmixup.readthedocs.io
Apache License 2.0
609 stars 61 forks source link

Support New Methods and More Features (V0.2.3) #6

Closed Lupin1998 closed 2 years ago

Lupin1998 commented 2 years ago

Supporting new methods:

  1. Provide README and update configs for self-supervised learning methods and supervised image classification methods.
  2. Support new Masked Image Modeling (MIM) methods (A2MIM, CAE).
  3. Support new Fine-tuing method (HCR).

Supporting new features:

  1. Support the online document of OpenMixup (built on Read the Docs).
  2. Support more regression losses (Focal L1/L2 loss, Balanced L1 loss, Balanced MSE loss).
  3. Support more regression metrics (regression errors and correlations) and the regression dataset.
  4. Support more reweight classification losses (Gradient Harmonized loss, Varifocal Focal Loss) from MMDetection.

Fix bugs:

  1. Fix the bug of DropPath (using stochastic depth rule) in ResNet for RSB A1/A2 training settings.
  2. Refactor code structures of openmixup.models.utils and support more network layers.
Lupin1998 commented 2 years ago

OpenMixup v0.2.3 is now released, which supports new methods and features. Configs and readme of both supervised and self-supervised methods are available.