search

facebookresearch / mmf

A modular framework for vision & language multimodal research from Facebook AI Research (FAIR)
https://mmf.sh/
Other
5.5k stars 939 forks source link

Annotation files used for calculating "bottom-up" features of OKVQA dataset #1042

Open ifmaq1 opened 3 years ago

ifmaq1 commented 3 years ago

❓ Questions and Help

The first step of KRISP model before VQA is to calculate visual features of OKVQA dataset. For this Bottom-up features are used probably from https://github.com/peteanderson80/bottom-up-attention .

Can you please tell, what annotation files are used for calculating the visual features of OKVQA dataset or train and validation split?

Also, please let me know do we need to calculate image features for whole coco dataset in order to calculate okvqa dataset's features? As, okvqa dataset only has 15k images but i am unable to find only those images anywhere.

apsdehal commented 3 years ago

@KMarino If you get time can you look into this? My guess is that https://github.com/facebookresearch/mmf/blob/master/tools/scripts/features/extract_features_vmb.py