Authors: Manuel Rota

Description: Text Localization using a FCN (VGG as a base) which is trained to predict bounding boxes area and bounding boxes divisions when two are too close.