Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

how it perfrom in AVA dataset? #7

Closed
accountcwd opened this issue Jan 14, 2020 · 11 comments
Closed

how it perfrom in AVA dataset? #7

accountcwd opened this issue Jan 14, 2020 · 11 comments

Comments

@accountcwd
Copy link

accountcwd commented Jan 14, 2020

thanks for open source code.

and i wonder can YOWO run on the AVA dataset?

@okankop
Copy link
Collaborator

okankop commented Jan 14, 2020

For AVA dataset, we need to formulate a new loss function since it is a multi-person multi-label dataset. We are currently working on it!

@accountcwd
Copy link
Author

accountcwd commented Jan 14, 2020

oh, that means your model is trained with single-person label, but when in infer phase it is able to output multi-person result?
I do not know if I have it wrong.

can i infer a large photo which contain many people?

@wei-tim
Copy link
Owner

wei-tim commented Jan 14, 2020

@accountcwd
In datasets UCF101-24 and J-HMDB-21, every single person has only one action label, but within an image, there can be several persons. As for AVA, it's not totally same, for each person might have two or more action labels.

You can use a large image containing several people, for each person there will be a single action prediction, like fencing in our examples.

@accountcwd
Copy link
Author

@accountcwd
In datasets UCF101-24 and J-HMDB-21, every single person has only one action label, but within an image, there can be several persons. As for AVA, it's not totally same, for each person might have two or more action labels.

You can use a large image containing several people, for each person there will be a single action prediction, like fencing in our examples.

Got it ! Thanks for detailed answer
@wei-tim @okankop

@dagongji10
Copy link

For AVA dataset, we need to formulate a new loss function since it is a multi-person multi-label dataset. We are currently working on it!

I have read your nice work for UCF101-24 dataset,and I wonder when will you finish the work for AVA ? Will you open source the AVA pretrained model ?
If possible, could you give me an approximate date.

@okankop
Copy link
Collaborator

okankop commented Jan 16, 2020

Hi @dagongji10, unfortunately we cannot give an approximate date now. Once we have YOWO working on AVA dataset, we would of course opensource pretrained models.

@okankop
Copy link
Collaborator

okankop commented Feb 8, 2021

AVA results are coming soon! 🎆

Stay tuned!

@irvingzhang0512
Copy link

AVA results are coming soon! 🎆

Stay tuned!

Looking forward to your work. I'm planning to reproduce YOWO with mmaction2 someday.

@tomatowithpotato
Copy link

AVA results are coming soon! 🎆

Stay tuned!

Will you launch yowo v2.0 in the future???????
looking forward to your work!!!

@okankop
Copy link
Collaborator

okankop commented Mar 29, 2021

YOWO is extended for AVA dataset! Please check out the updated repo!

@okankop
Copy link
Collaborator

okankop commented Mar 29, 2021

@tomatowithpotato yes I am working on YOWOv2, which will be faster and stronger!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

6 participants