EMPOWER COUNTERFACTUAL THINKING VIA CONTRASTIVE LEARNING FOR ROBUST VISUAL QUESTION ANSWERING

The code will be available soon(after accepted)!

Paper:

EMPOWER COUNTERFACTUAL THINKING VIA CONTRASTIVE LEARNING FOR ROBUST VISUAL QUESTION ANSWERING

Training

All the args for running our code is preset in the main.py.

Run

CUDA_VISIBLE_DEVICES=0 python main.py

Our trained models on VQA-CP v2 can be obtained from BaiduYun：link will be here

Citation