chojw / genb

Generative Bias for Robust Visual Question Answering ( CVPR 2023 )

Generative Bias for Robust Visual Question Answering

This repo contains the PyTorch code release for our paper GenB: Generative Bias for Robust Visual Question Answering (CVPR 2023).

Prerequisites

Please make sure you are using a NVIDIA GPU with Python==3.7.1 and about 100 GB of disk space.

Install all requirements with pip install -r requirements.txt

Data Setup

Download UpDn features from google drive, which is the link from this repo, into /data/detection_features folder

Download questions/answers for VQA v2 and VQA-CP2 by executing bash tools/download.sh

Preprocess process the data with bash tools/process.sh

Training

Run python main.py to run GenB.

Evaluating

Run python eval.py --load_path DIRNAME to evaluate your model.

For the best performing model, you can download our best performing model from here and run it.

Acknowledgements

This repo contains code largely modeified from here.

Citation

If you find this code useful, please cite our paper:

@inproceedings{cho2023genb,
	title={Generative Bias for Robust Visual Question Answering},
	author={Cho, Jae Won and Kim, Dong-Jin and Ryu, Hyeonggon and Kweon, In So},
	booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
	year={2023}
}