Could you please provide more examples to do inference on the different tasks in the paper?

[buaalyx]

Such as temporal grounding on QVHighlight and Charade-STA

[dongdk]

+1

[arushirai1]

Same question here. I have tried getting an output in frames or seconds and both seem to perform poorly.

[Shuaicong97]

Any update? @shepnerd @buaalyx I tried to use InternVideo2/demo, since they're using the pretrained Bert, the feature dim is 512.

[Magmanat]

+1