Closed scottclowe closed 9 years ago
Can now pass a "n" value to "get_ORB_keypoints()" representing the max no of features we would like. In addition, we can change the "patchSize" parameter of the ORB detector in order to get keypoints for small images (i.e. by deceasing patchSize). BUT When describing the keypoints, we have to make sure that the descriptor is initialised with the same value for "patchSize".
"BUT When describing the keypoints, we have to make sure that the descriptor is initialised with the same value for "patchSize". --> We do this by passing **kwargs. (cheers Scott)
Use OpenCV to detect oriented BRIEF keypoints and then describe them.
http://docs.opencv.org/trunk/modules/features2d/doc/feature_detection_and_description.html#orb
Must ensure parameters are suitable so EVERY image in BOTH train and test datasets have an appropriate number of keypoints detected.