Face Localization


In this project we implemented a simple Face Localization (this means finding faces in images) based on OpenCV and the Deep Learning framework Caffe.

The code was tested only on Linux. A makefile is included. However you surely have to adapt this file to your needs. The basic ideas and problems behind this project are described here . This project was created as a by-product of the german ebook “Deep Learning – Part 1“.

The underlying code is very simple and rudimentary. Therefore there is still much to improve in terms of the C/C ++ code as well as the Caffe network used.

Downloadlink of C/C++ Source: FaceNet C/C++

Downloadlink of Python Source (same algorithms as C/C++ but in Python) : FaceNet Python

Downloadlink of trained Neural Net (caffemodel): FaceNet Caffemodel

Some results (Red rectangles mark faces recognized by the software;
These recognized areas do not necessarily have to be real faces):

val_c_1 val_c_3 val_c_5 val_c_6
inv_c_1 inv_c_3 inv_c_4 val_w_8