jimmy-dq / dropmae Goto Github PK

Congrats for such a great work! May I ask why the AiATrack results in Table 2 of your arXiv verison are so low? I am happy to provide the raw results or any other kinds of help if you meet trouble in reproducing AiATrack. Thanks!

...

Hi, What type of gpu is recommended to train the model?

Your work is exciting!!!

Dear Authors:
Thanks for your great work!
I have downloaded the Kinetics 700-2020 which is provided in this link.
However, i find there are just urls in *.json rather than video files after decompressing it.
Could you provide some scripts to download videos files from the urls, splitting them according to *.csv?
I am very grateful if you provide further instructions about Kinetics dataset!

I'm impressed by your great work!

When loading the model using a checkpoint, the keys do not match.

Could you share full checkpoint that includes decoder?

I noticed someone earlier pretrain the vit using masked autodecoder on video data: eg. VideoMAE and VideoMAEv2. Do you think that pretrained vit could also be finetuned in downstream vot and vos tasks. And it has the similiar spirit with your work?

Thank you so much for sharing the code of this work!

Do you mind releasing the pretrained ViT checkpoint?

Many thanks!

Thank you so much for sharing the checkpoint and codes of pretraining. And do you mind releasing the finetune codes for vot and vos? Thanks a lot.