method: Papago OCR (PixelLink+)2019-04-26
Authors: Chankyu Choi, Youngmin Yoon, Hong-Seok Lee, Minyoung Jun
Description: Our detector is based on PixelLink, modified w/ ResNet + SE. We changed some parameters from the original code such as pixel and link threshold. Also, our input image size is 1024x1024. We used only LSVT full dataset to train our models. (Papago OCR team, NAVER corp.)