Training data for cup detection

asked 2017-02-06 00:53:13 -0500

Akhilesh gravatar image

I am new in opencv. I am trying to make cup detector like face detector in opencv. But I see the positive samples of face detector, there is particular pattern of eyes, nose and mouth, so there is, somewhat all, positive samples have same pattern. But cups don't have the same pattern on it. Some cups are plan and some have graphics on it. So suggest me any idea how to take samples and implement it.

edit retag flag offensive close merge delete



Generally speaking, OpenCV does not support training models complex enough for such a task. You are better off turning to neural network libraries such as TensorFlow and perhaps even decision trees/forests. Decide what kind of data you have (RGB, gray, depth) and go from there. Are you interested in per-pixel classification? In image classification? In a bounding box? In the pose of the cup?

Also, think about Plato vs. Aristotle. How is it that you can recognize a cup, from a likeness to a perfect form or from all known examples?

Der Luftmensch gravatar imageDer Luftmensch ( 2017-02-06 19:31:06 -0500 )edit

Thank you Der Luftmensch. I think it is a good suggestion for me to try TensorFlow.

Akhilesh gravatar imageAkhilesh ( 2017-02-07 01:54:17 -0500 )edit

Hi, can you please refer any link or tutorial to train my data of cup detection.

Akhilesh gravatar imageAkhilesh ( 2017-02-13 06:41:11 -0500 )edit

Thank you Der Luftmensch

Akhilesh gravatar imageAkhilesh ( 2017-02-14 03:28:27 -0500 )edit