Learning OpenCV 4 Computer Vision with Python 3 - Third Edition

Applying BoW to computer vision

We are by now familiar with the concepts of features and descriptors. We have used algorithms such as SIFT and SURF to extract descriptors from an image's features so that we can match these features in another image.

We have also recently familiarized ourselves with another kind of descriptor, based on a codebook or dictionary. We know about an SVM, a model that can accept labeled descriptor vectors as training data, can find an optimal division of the descriptor space into the given classes, and can predict the classes of new data.

Armed with this knowledge, we can take the following approach to build a classifier:

1. Take a sample dataset of images.
2. For each image in the dataset, extract descriptors (with ...