Mucko

implementation for Mucko: Multi-Layer Cross-Modal Knowledge Reasoning for Fact-based Visual Question Answering

Requirements

pytorch == 1.2.0
dgl == 0.4.3

Train

python train.py --dataset okvqa --validate --save-dirpath save_dir --gpu-ids 0

Reference

@inproceedings{zhu1020mucko,
  author    = {Zihao Zhu and
               Jing Yu and
               Yujing Wang and
               Yajing Sun and
               Yue Hu and
               Qi Wu},
  title     = {Mucko: Multi-Layer Cross-Modal Knowledge Reasoning for Fact-based
               Visual Question Answering},
  booktitle = {Proceedings of the International Joint Conference on Artificial Intelligence},
  pages     = {1097--1103},
  year      = {2020}
}