how to use ORB features with scikit learn
I am developing an image classification program for classifying food images using machine learning algorithms in python. Currently I have been using the histograms as the feature for classification but now I want to make use of ORB features.Is it possible to use the features with scikit learn ? I am using Anaconda and my python version is 3.6. can somebody guide me? Please do not direct me to SIFT or SURF resources as they are patented.
since every image will have varying number of keypoints / descriptors, and you need a fixed descriptor size for any kind of machine learning, you will need to cluster the descriptors e.g. using BagOfWords.
so: find out, if scikit has any ways to cluster binary features
(this is somewhat off-topic here)
btw, you can still use opencv's BoW methods using AKAZE with KAZE_UPRIGHT descriptors (they are float, and not patented)