Authors: Shuai Ren

Description: We use the psenet's code, which is a method who use progressive scale expansion algorithm, it can effectively distinguish the text instances lying closely. The network backbone we used is resnet152. We change the learning rate decline to warmup.
Art train dataset is the only train dataset.

Author:
Shuai Ren, YaFei Wen
Tel:
+86 15008276006
Email:
shuai.ren@vivo.com
Company:
VIVO AI Lab

作者：
任帅,文亚飞
联系电话：
+86 15008276006
电子邮箱：
shuai.ren@vivo.com
公司：
VIVO深圳AI研究院