Learning to Contrast the Counterfactual Samples for Robust Visual Question Answering
The source code for our paper Learning to Contrast the Counterfactual Samples for Robust Visual Question Answering published in EMNLP 2020. This repo contains code modified from CSS-VQA, Many thanks for their efforts.
Prerequisites
Make sure you are on a machine with a NVIDIA GPU and Python 2.7 with about 100 GB disk space.
h5py==2.10.0
pytorch==1.1.0
Click==7.0
numpy==1.16.5
tqdm==4.35.0
Data Setup
All data preprocess and set up please refer to bottom-up-attention-vqa
- Please run the script to download the data.
bash tools/download.sh
Training
All the args for running our code is preset in the main.py.
Run
CUDA_VISIBLE_DEVICES=0 python main.py
to train a model
Testing
Run
CUDA_VISIBLE_DEVICES=0 python eval.py --dataset [] --debias [] --model_state []
to eval a model
Citation
If you find this paper helps your research, please kindly consider citing our paper in your publications.
@inproceedings{liang2020learning,
title={Learning to Contrast the Counterfactual Samples for Robust Visual Question Answering},
author={Liang, Zujie and Jiang, Weitao and Hu, Haifeng and Zhu, Jiaying},
booktitle={Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)},
year={2020}
}