Dear Experts,
I would like to prepare software that allows me to capture a gesture. It will be very useful in my further project sucha as a game or smart white board.
The algorithm flow is quite obvious.
- get a frame
- image preprocessing ( binarization, application of some filters, erosion, dilatation etc.)
- feature extraction
- Neural network training (training phase) or optional use NN to detect gesture (run phase)
My question is how to perform feature extraction. how can get the vector of feature from image.
As a classifier I would like to use feed forward NN not any CNN. I would like to perform such operation without deep learning that mean in supervised way.I don't want to put into network input layout whole image.
Can you give me any hint how to apply Machine learning for such problems?