Authors: PINGAN Tech Computer Vision Technology Group

Description: The text localization was implemented by pixel link, a kind of instance segmentation method. Vgg is used as backbone and the whole network was trained in multi-scale.