CCD

Official PyTorch implementation of the paper "Contextual Debiasing for Visual Recognition with Causal Mechanisms"

A Simple but Effective Baseline for Multi-label Classification

Requirements:

• PyTorch >= 1.7.0
• torchvision >= 0.5.0
• randaugment
• pycocotools
• opencv-python
• pillow

Training

To train a model, run train.py with the desired model architecture and the path to the dataset. For example, to train with full model on MS-COCO with resnet101 backbone:

CUDA_VISIBLE_DEVICES=0,1 python3 -u train.py --batch-size=128 --image-size=448 --lr=1e-4 --backbone=resnet101  --loss=focal --use_intervention=True --use_tde=True --stop_epoch=5

to train with full model on MS-COCO with Swin-B backbone:

CUDA_VISIBLE_DEVICES=0,1 python3 -u train.py --batch-size=32 --image-size=384 --lr=5e-5 --backbone=swim_transformer --loss=halfasl --use_tde=True --use_intervention=True --stop_epoch=2

Citation

If you find our paper or this project helps your research, please kindly consider citing our paper in your publications.

@inproceedings{liu2022contextual,
  title={Contextual Debiasing for Visual Recognition With Causal Mechanisms},
  author={Liu, Ruyang and Liu, Hao and Li, Ge and Hou, Haodi and Yu, TingHao and Yang, Tao},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={12755--12765},
  year={2022}
}