jnhwkim/ban-vqa

train36_imgid2idx.pkl file

atg93 opened this issue · 1 comments

atg93 commented

Hi, thank you for sharing your code. I was wondering what does data/train36_imgid2idx.pkl contain exactly ?

Image index mapping for the bottom-up attention features where the fixed 36-box features per image are extracted. By the way, we used 10-100 boxes, though.