Implementation of the paper "Text Detection and Character Recognition in Scene Images with Unsupervised Feature Learning" from Andrew Ng et. al.
The bootstrap.sh script gets you started by unpacking the training data and running the different stages of the project.
- not used but possible: http://www.robots.ox.ac.uk/~vgg/data/text/
Extracts all patches used for training. File patch_extraction.py Note: The extracted text patches should have a size such that a 32x32 image roughly corresponds to the size of one character
File feature_learning contains the k-means algorithm which learns the dictionary wich is used for the feature extraction.
...
...