the dataset: Kinetics-400 dataset link

[BryanWu404]

I noticed that the link to the Kinetics-400 dataset mentioned in the documentation is currently unavailable (broken link).

Could you please update the URL or provide an alternative source?

Thanks.

[ssx263594]

Sorry, I am not the original author of this file. I am just using it for study and reproduction purposes. I only used the UCF101 dataset to verify the model, and the results were quite similar to the original. Other datasets might perform better due to their larger size, but I do not have the source for Kinetics-400 (K400). I suggest you contact the original author regarding that。

…

---- Replied Message ---- | From | ***@***.***> | | Date | 12/18/2025 19:56 | | To | ***@***.***> | | Cc | ***@***.***> | | Subject | [Visual-AI/FROSTER] the dataset: Kinetics-400 dataset link (Issue #18) | BryanWu404 created an issue (Visual-AI/FROSTER#18) I noticed that the link to the Kinetics-400 dataset mentioned in the documentation is currently unavailable (broken link). Could you please update the URL or provide an alternative source? Thanks. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[BryanWu404]

Thanks for the reply. In that case, I will use UCF101 to reproduce the results instead.

[BryanWu404]

Sorry to bother you again. I attempted to reproduce the results using UCF101 but encountered persistent failures due to pytorchvideo version incompatibilities and hardware resource constraints (OOM).

Could you please share the inference logs or the saved output tensors used to generate the visualizations (like Fig. 4)?

No problem if that's not feasible. Thanks! I appreciate your help.