This notebook uses a deep convolutional neural network (CNN) to extract features from input images. It follows the approach described in [1] with modifications inspired by the OpenFace project. Keras is used for implementing the CNN, Dlib and OpenCV for aligning faces on input images. Face recognition performance is evaluated on a small subset of the LFW dataset which you can replace with your own custom dataset e.g. with images of your family and friends if you want to further experiment with this notebook. After an overview of the CNN architecure and how the model can be trained, it is demonstrated how to:
Source: Jupyter Notebook Viewer