Learning Data Mining with Python - Second Edition

Splitting the image into individual letters

Our CAPTCHAs are words. Instead of building a classifier that can identify the thousands and thousands of possible words, we will break the problem down into a smaller problem: predicting letters.

Our experiment is in English, and all uppercase, meaning we have 26 classes to predict from for each letter. If you try these experiments in other languages, keep in mind the number of output classes will have to change.

The first step in our algorithm for beating these CAPTCHAs involves segmenting the word to discover each of the letters within it. To do this, we are going to create a function that finds contiguous sections of black pixels in the image and extract them as subimages. These are (or at least ...