open-mmlab
Real-time detection with mmaction2？
Branch
main branch (1.x version, such as v1.0.0, or dev-1.x branch)
Prerequisite
- [X] I have searched Issues and Discussions but cannot get the expected help.
- [X] I have read the documentation but cannot get the expected help.
- [X] The bug has not been fixed in the latest version.
Environment
Operating System: Win11 mmaction2 version: latest GPU/CPU: gpu
Describe the bug
As a beginner in using mmaction2, I am currently exploring the workflow for skeleton action recognition. I noticed that the suggested process involves using "mmdetection+mmpose+mmaction". However, when I tried running the demo, it appeared to be quite slow. Each step, from mmdetection to mmpose to mmaction, takes a significant amount of time to process.
Reproduces the problem - code sample
No response
Reproduces the problem - command or script
No response
Reproduces the problem - error message
No response
Additional information
No response
Hi, @heylary It's reasonable, the detection and pose models used here are not efficient enough. Actually, we're going to release a real-time skeleton-based gesture recognition project, which uses real-time detection and pose models, you could try it later.
Hi @cir7, I am wondering when are you going to release it and what will its name be? Also, what is the size of the model gonna be and can it be run on edge device? Thanks!
I have tried using this for real-time recognition and it works OK, but the results are not very good。 https://github.com/open-mmlab/mmaction2/blob/main/demo/README.md#webcam-demo
