method: rest_submit_task2_with_pretrain_0_6000_new2023-03-21

Authors: Zhong Humen

Description: The recognition model is trained on a mixed dataset with ReST training set and synthetic images. The detection model is from taks1.