Ultra Light OCR Competition Technical Report

TL;DR

This paper reports on the Ultra Light OCR Competition focusing on Chinese scene text recognition within a 10M model size limit, proposing effective methods that achieved second place with 81.7% accuracy.

Contribution

It introduces a general and effective approach for Chinese scene text recognition balancing model size and accuracy, validated through competitive results.

Findings

Achieved 81.7% accuracy in TestB dataset

Developed a method balancing model scale and recognition performance

Secured second place among over 100 teams

Abstract

Ultra Light OCR Competition is a Chinese scene text recognition competition jointly organized by CSIG (China Society of Image and Graphics) and Baidu, Inc. In addition to focusing on common problems in Chinese scene text recognition, such as long text length and massive characters, we need to balance the trade-off of model scale and accuracy since the model size limitation in the competition is 10M. From experiments in aspects of data, model, training, etc, we proposed a general and effective method for Chinese scene text recognition, which got us second place among over 100 teams with accuracy 0.817 in TestB dataset. The code is available at https://aistudio.baidu.com/aistudio/projectdetail/2159102.