search

open-mmlab / mmaction2

OpenMMLab's Next Generation Video Understanding Toolbox and Benchmark
https://mmaction2.readthedocs.io
Apache License 2.0
4.04k stars 1.2k forks source link

Do we now support the method of Skeleton-based Spatio-Temporal Action Detection and Action Recognition #2823

Open SCsleepy opened 2 months ago

SCsleepy commented 2 months ago

The doc issue

Hello， I saw an example on the homepage titled 'Skeleton based Spatial Temporary Action Detection and Action Recognition Results on Kinetics-400'. May I ask if Skeleton based Spatial Action Detection and Action Recognition is currently supported? I couldn't find the corresponding inference script and configuration file. Because the current skeleton based behavior recognition assumes that a video is a behavior, I hope to achieve multiple individuals in a video given their respective behaviors. I hope to receive your help。 image

Suggest a potential alternative/fix

No response

SCsleepy commented 2 months ago

I would also like to inquire about the format of my dataset if I want to implement Skeleton based Spatial Temporary Action Detection and Action Recognition.Thank you for your help

liukp557 commented 2 months ago

Hello, I have the same problem, I would like to use a bone point based behavior recognition model for inference to recognize multiple character actions in a video separately, is there a relevant inference script provided?

Shifiter commented 2 weeks ago

I have the same problem, Skeleton-based Spatio-Temporal Action Detection seems to have not clearly stated which model was used and its training method.

mbotsu commented 4 days ago

Video Structuralize Demo So, this kind of detection is possible, but does it match what is being said?

https://github.com/open-mmlab/mmaction2/blob/main/demo/README.md#video-structuralize-demo

https://github.com/open-mmlab/mmaction2/assets/33709712/7434c048-459b-485a-b2b9-fd75a3ac5fb3