open-mmlab
Do we now support the method of Skeleton-based Spatio-Temporal Action Detection and Action Recognition
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。
Suggest a potential alternative/fix
No response
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
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?
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.
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
