First of all thank you for this great work, it helped me!
I expect to be able to use YOLOPOSE for insecure behavior sorting
Motivation
I always get frustrated when I don't know how to proceed to the next step of human action classification
Pitch
I expect to be able to detect unsafe passenger behavior on escalators and run on some CPU devices rather than GPU devices. YOLOPOSE is small and fast enough to detect the key points of human bones on some CPU devices, and I want to know how to complete the classification of behaviors based on YOLOPOSE, because I need to alert the unsafe behaviors. For example, if a person falls on an escalator, I expect to be able to detect the key bone point of the fall, recognize that the person is in a fallen state, and then remind others to help him.
Alternatives
I only know that svm and ST-GCN seem to be able to realize action classification, but I don't know anything about this aspect at all. I hope you can give me a suggestion on which network to combine YOLOPOSE to realize action recognition, and this work can be fast enough
π Feature
First of all thank you for this great work, it helped me!
I expect to be able to use YOLOPOSE for insecure behavior sorting
Motivation
I always get frustrated when I don't know how to proceed to the next step of human action classification
Pitch
I expect to be able to detect unsafe passenger behavior on escalators and run on some CPU devices rather than GPU devices. YOLOPOSE is small and fast enough to detect the key points of human bones on some CPU devices, and I want to know how to complete the classification of behaviors based on YOLOPOSE, because I need to alert the unsafe behaviors. For example, if a person falls on an escalator, I expect to be able to detect the key bone point of the fall, recognize that the person is in a fallen state, and then remind others to help him.
Alternatives
I only know that svm and ST-GCN seem to be able to realize action classification, but I don't know anything about this aspect at all. I hope you can give me a suggestion on which network to combine YOLOPOSE to realize action recognition, and this work can be fast enough
Additional context
Look forward to your suggestions! thank you