KPS Loss: Key Point Sensitive Loss for Long-tailed Visual Recognition

This is the source code for our TPAMI paper: Key Point Sensitive Loss for Long-tailed Visual Recognition.
This version is a demo of how to use KPS loss. The version that supports more datasets is in the works and is coming soon.

CIFAR10

$ python cifar_train_backbone.py --arch resnet32 /
                                 --dataset cifar10 --data_path './dataset/data_img'/
                                 --loss_type 'KPS' --train_rule 'GA'/
                                 --imb_factor 0.01/ 
                                 --batch_size 64 --learning_rate 0.1 

To do list:

• Support Cifar10/100-LT dataset
• Support imageNet-LT
• Support iNaturalist2018
• More loss functions
• Separate configuration files for easier execution
• Some other minor performance improvements

Citation

@article{Li2022Long,
  author    = {Mengke Li, Yiu{-}ming Cheung, Zhikai Hu},
  title     = {Key Point Sensitive Loss for Long-tailed Visual Recognition},
  journal   = {{IEEE} Transactions on Pattern Analysis and Machine Intelligence},
  volume    = {},
  number    = {},
  pages     = {in press},
  publisher = {IEEE},
  year      = {2022},
  doi       = {10.1109/TPAMI.2022.3196044},
}

Other Resources of long-tailed visual recognition

Awesome-LongTailed-Learning

Awesome-of-Long-Tailed-Recognition

Long-Tailed-Classification-Leaderboard

BagofTricks-LT

GCL: Long-tailed Visual Recognition via Gaussian Clouded Logit Adjustment

Connection

If you have any questions, please send the email to Mengke Li at: [email protected].