search

ItemZheng / KDDAug

[ECCV2022] Rethinking Data Augmentation for Robust Visual Question Answering
10 stars 0 forks source link

Google drive for faster rcnn features and *hintscore.json is invalid #2

Closed alice-cool closed 1 year ago

alice-cool commented 1 year ago

Download faster rcnn features Download feature1.zip and feature2.zip from Google Drive, then unzip and merge them into data/rcnn_feature/.

alice-cool commented 1 year ago

Download extra data to train CSS (ID & OOD Teacher in KDDAug) Download *hintscore.json files from here, and move them to data folder.

alice-cool commented 1 year ago

image

ItemZheng commented 1 year ago

The repo CSS-VQA updated download url. And you can download here: https://mega.nz/folder/0JBzGBZD#YGgonKMnwqmeSZmoV7hjMg