Creating features using visual codebook and vector quantization

In order to build an object recognition system, we need to extract feature vectors from each image. Each image needs to have a signature that can be used for matching. We use a concept called visual codebook to build image signatures. This codebook is basically the dictionary that we will use to come up with a representation for the images in our training dataset. We use vector quantization to cluster many feature points and come up with centroids. These centroids will serve as the elements of our visual codebook. You can learn more about this at

Before you start, make sure that you have some training images. ...

Get Python: Real World Machine Learning now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.