NExT-GQA / code /SeViLA /README.md
jinyoungkim's picture
Add files using upload-large-folder tool
52d2b19 verified
|
Raw
History Blame Contribute Delete
473 Bytes

Self-Chained Image-Language Model for Video Localization and Question Answering

Please follow the official code to reproduce the results on NExT-GQA. Afterwards, you can use the tools provided here to convert the predicted frame ids into timestamps for evaluation. To run the code, you need to copy tools into sevila and change the file paths accordingly.

python fid2seconds.py
python split_qa_los.py