Awesome

A Conservative Approach for Unbiased Learning on Unknown Biases (CVPR2022)

@inproceedings{jeon2022conservative,
  title={A Conservative Approach for Unbiased Learning on Unknown Biases},
  author={Jeon, Myeongho and Kim, Daekyung and Lee, Woochul and Kang, Myungjoo and Lee, Joonseok},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={16752--16760},
  year={2022}
}

Abstract

Although convolutional neural networks (CNNs) achieve state-of-the-art in image classification, recent works address their unreliable predictions due to their excessive dependence on biased training data. Existing unbiased modeling postulates that the bias in the dataset is obvious to know, but it is actually unsuited for image datasets including countless sensory attributes. To mitigate this issue, we present a new scenario that does not necessitate a predefined bias. Under the observation that CNNs do have multi-variant and unbiased representations in the model, we propose a conservative framework that employs this internal information for unbiased learning. Specifically, this mechanism is implemented via hierarchical features captured along the multiple layers and orthogonal regularization. Extensive evaluations on public benchmarks demonstrate our method is effective for unbiased learning.

Categorization of Data Bias Problems with an example of gender prediction biased in hair length. The most limited case, unknown bias, is what we address in this paper.

The architecture of the proposed model, UBNet

The architecture of the proposed model, UBNet. UBNet takes the hierarchical features captured by the base model as input. The Trans-layers set all the <img src="https://render.githubusercontent.com/render/math?math=\mathbf{h}_l"> be the same size. Then, all the concatenated <img src="https://render.githubusercontent.com/render/math?math=\mathbf{g}_l"> activated through the Ortho-block in which Ortho-Conv and Ortho-Trans layers encode multi-variant features. From the output of the Ortho-block, each classifier outputs confidence scores for each low-to-high feature. They are averaged for the final prediction. In this figure, we use L = 5 for simplicity.

Experiments

Requirements

Dependencies can be installed via anaconda.

python>=3.7
pytorch==1.9.0
torchvision==0.10.0
numpy=1.18.5
adamp=0.3.0
opencv-python=4.5.3.56
cudatoolkit=11.1.74

Dataset Preparation

CelebA-HQ

Download the CelebA-HQ and CelebA in the links as below.

• https://github.com/switchablenorms/CelebAMask-HQ
• https://mmlab.ie.cuhk.edu.hk/projects/CelebA.html (In-The-Wild Images)

We expect the directory sturcture to be the following

/data/
  celebA/
    celebA/
      img_celeba/     # CelebA images in wild
    celebA-HQ
      CelebA-HQ-img/  # CelebA-HQ images

Usage

Pretrained weights

Download the weights in https://drive.google.com/drive/folders/1_Dkr4CAPxWHbkOU7PIV3gbg9-oqotWRI?usp=sharing

Training

base model

python main.py -e celeba --is_train --is_valid --cuda --imagenet_pretrain --data CelebA-HQ --n_class 2 --lr 0.0001 --max_step 20 --gpu 1 --model vgg11 --lr_decay_period 10 --lr_decay_rate 0.1

UBNet

python main.py -e celeba_ubnet --is_train --is_valid --ubnet --cuda --use_pretrain True --checkpoint celeba/checkpoint_step_19.pth --data CelebA-HQ --n_class 2 --lr 0.0001 --max_step 20 --lr_decay_rate 0.1 --lr_decay_period 10 --model vgg11 --gpu 1

Evaluation

Note that we have reported the average of 3 evaluations. The uploaded weight is from one of the 3 experiments; ACC(EB1) 99.29%, ACC(EB2) 58.74%, and hence ACC(Test) 79.02%.