I am looking at detecting different physical objects using SURF algorithm. Is this good for real time object detection? OR should I look for AAM/Haar?

My work flow is as follows:

Teach an object, say a pet bottle by capturing the bottle image from different sides. Generate keypoints and feature descriptors for it. Now when I test it with an object from live camera feed it should check for match using SURF/FAST. Based on a threshold matching value I decide whether the object is the right one.

Expecting your valuable opinion on this.

this really sounds like a good job for a machine learning algorithm.

maybe give tensorFlow a try:


or try some of the OF addons for machine learning:


Thanks a lot for your feedback. I am trying out various libraries now. Got stuck in ofxFacetracker by kyleMacDonald. Have some issues building it.

I have raised a separate topic on this.