method: NJU-ImagineLab(v3)2019-06-03

Authors: Wenhai Wang, Yao Xiao, Xiaoge Song

Description: It is a two-stage end-to-end detection framework, we use Mask R-CNN as our baseline and adapt it for Multi-lingual scene text detection. Our approach first predicts instance segmentation for texts and then extracts quadrilateral bounding boxes from segmentation maps.
ImagineLab, Nanjing University