Reproducing of TAD

[gyusik19]

Awesome Work! Do you have plans for sharing the reproduction of Temporal Action Detection Task?

[congee524]

Hello! For the TAD task, we use a two-stage approach: after extracting features with VideoMAE V2-g, we trained using the ActionFormer codebase without making any modifications to the method. In the next few days, we will release the features of THUMOS’14 and FineAction extracted by VideoMAE V2-g. Please stay tuned for updates.

[gyusik19]

Thank you very much for your reply!! will VideoMAE V2-g model also be available for extracting features of my own dataset?

[congee524]

Yes! We will also release the giant model :p

[gyusik19]

Thank you for your reply :)

[congee524]

Thank you for your reply :)

Giant model is available :-D

[Value-Jack]

if possible，could you please extract the features of ActivityNet1.2 extracted by VideoMAE V2-g and share it in the future? thank you very much!