Recently, I was learning about motion capture and facial expression recognition, both of which involve machine learning. Overall, I am confused, and I admonish myself that I will not eat fat at all. Perseverance is only effective.
Contacted the two popular libraries openpose and openface on the market.
openpose is good at body gesture recognition (both with facial recognition function but rough), openface is good at facial recognition (not including body). The two frameworks have different application areas, the overall effect is good, and the recognition quality is also very high.
Various problems were encountered in the process of use, and some of them were machine environment problems. Openpose has relatively high requirements on machine hardware, especially when it comes to the acceleration part of the graphics card, and the environment configuration also has certain requirements. There is also a part of the network download dependent packages and models, some need to over the wall, and some connections are slow.
Share yourself with the downloaded framework (you can run the demo), so you do n’t have to waste time on the environment.
