Development Blog Init
Of initial interest is pose estimation. A person’s body language can be analyzed for pre-assaultive behavior cues such as blading (angling towards a target with dominant foot back), pacing, etc.
To start, the Deep High-Resolution Representation Learning for Human Pose Estimation (CVPR 2019) system was successfully installed on a GPU-based AWS EC2 instance, and a handful of videos processed with it.
The slow performance, ie, low FPS rate will need to be addressed with better hardware, faster software, or likely a combination of both, but for proof-of-concept, it will do for now.