Description: We are from the Group of Computer Vision, 4paradigm. Our approach mainly follows the framework of two-stage detection network. Use icdar2017mlt 9000 train images and icdar2019mlt 9000 train images.
Description: We are from the Department of Computer Vision, Baidu Inc. Our approach follows the framework of LOMO and mainly upgrades the backbone network. For more details about LOMO, please refer to the paper: "Look More Than Once: An Accurate Detector for Text of Arbitrary Shapes", accepted by CVPR2019. (https://18.104.22.168/abs/1904.06535)