MILVLG / mcan-vqa

Deep Modular Co-Attention Networks for Visual Question Answering
Apache License 2.0
442 stars 88 forks source link

Are the provided bottomup feature same as the it in the original repo? #17

Closed Wangt-CN closed 5 years ago

Wangt-CN commented 5 years ago

Hi, thank you for sharing your code. However, I have found that the bottom up feature you provided is different with the default bottom up feature? Which is different with the up down model and BAN?