Feactures related to the selection of KNN about uav-gesture-control_python HOT 4 OPEN

You can check this part in the code. Actually, I used the dot product of two vector to distinguish different poses.

from uav-gesture-control_python.

You can check this part in the code. Actually, I used the dot product of two vector to distinguish different poses.

hi，I am very happy to receive your reply.
Since I have seen your pose diagram, I thought before that how can the parameters (（x2-x1),(y2-y1）/（|x2-x1| + |y2-y1|） ) be distinguished, so I would like to ask you.
thanks.

I have been working on pose classification based on key points recently. Do you have any suggestions?

from uav-gesture-control_python.

I have tried Euclidean distance and vector angle between the training vector and test vector. I found the vector angle has a better performance. The recognition accuracy is larger than 90%. I have no idea to improve this project. Nevertheless, if you want to try, deep learning can be a potential approach.

from uav-gesture-control_python.

Hello, sorry for the late reply. saw it very early actually, but really do not know why cannot reply, gradually delayed.
I really appreciate your prompt reply and many suggestions.

from uav-gesture-control_python.