An image recognition library using TensorFlow and Keras.
Inside tensorflow_scripts, there is a set of scripts for formatting and compressing the data, and training and testing a CNN for face recognition implemented using TensorFlow.
The dataset used comes from the Extended Yale Faces Database B, publicly available at: http://vision.ucsd.edu/content/extended-yale-face-database-b-b
The face_recognizer.py file allows the addition of image classes, and an artificial generation of data, using image translations and rotations. It also provides an easy way to build a neural network using the Keras.layers classes, as well as training and validating it using the images previously added.
A test of the proposed library can be found in the paper: http://ijoer.com/Paper-August-2017/IJOER-AUG-2017-25.pdf