In Defense of Grid Features for Visual Question Answering

Benchmark Model Rank Results
visual-question-answering-on-vqa-v2-test-stdSingle, w/o VLP#18overall: 74.16yes/no: 89.18number: 58.01other: 64.77
visual-question-answering-on-vqa-v2-test-stdX-101 grid features + MCAN#21overall: 72.71
visual-question-answering-on-vqa-v2-test-devX-101 grid features + MCAN#22Accuracy: 72.59