Authors: liu

Affiliation: liu

Description: Mainly use segmentation-based methods. A two-stage approach is used for end-to-end recognition. The detection part mainly uses DBNET, and the recognition part mainly uses transformer. And only use official data for training