Skip to content

Latest commit

 

History

History
27 lines (21 loc) · 637 Bytes

README.md

File metadata and controls

27 lines (21 loc) · 637 Bytes
Raw

ASCL: Adaptive Self-supervised Counterfactual Learning for Robust Visual Question Answering

Prerequisites

  • python 3.7
  • pytorch 1.9.1
  • pytorch-metric-learning 2.1.0
  • tdqm
  • Pillow
  • transformers

Download and preprocess the data

cd data 
python compute_softscore.py v2
python compute_softscore.py cp_v1
python compute_softscore.py cp_v2
cd ..

the data can also be obtained from BaiduYun(passwd:et10)

unzip rcnn_features.zip into data/rcnn/

the trained model can be obtained from BaiduYun(passwd:6eh9)