Tsinghua University News

Associate Professor Peng Liangrui’s research group wins in ICDAR 2017 Arabic Video Text Detection and Recognition Competition

Associate Professor Peng Liangrui’s research group wins in ICDAR 2017 Arabic Video Text Detection and Recognition Competition


Recently, the International Conference on Document Analysis and Recognition (ICDAR 2017) was held in Kyoto, Japan. The results of the first Arabic Video Text Detection and Recognition Competition were announced, and Associate Professor Peng Liangrui together with her research group from the Department of Electronic Engineering at Tsinghua University were champions in the seven sub-tasks of text detection, and in three of the seven sub-tasks of text recognition.

The ICDAR 2017 AcTiV competition award was presented to the winner by the competition organizer. From the right, the second is Dr. Peng Liangrui, and the third is PhD candidate Yan Ruijie.

ICDAR is the most important international conference in the field of Optical Character Recognition (OCR) organized by the International Association for Pattern Recognition (IAPR) since 1993. As there are increasing research interests on Arabic video OCR, ICDAR hosted the competition on Arabic video text detection and recognition (Arabic Text Detection and Recognition in Multi-resolution Video Frames, AcTiV) for the first time. The competition was mainly organized by the University of Fribourg in Switzerland and it included two tasks ---- text detection and text recognition. The samples used in the completion covered different font types and scales, various colors, different degrees of background complexity and low resolution. Participants included Tsinghua University, the Institute of Automation of the Chinese Academy of Sciences, and the University of Sfax (Tunisia), among others.

The award certificate in Arabic text detection.

The award certificate in Arabic text recognition.

The video text detection and recognition algorithms submitted by Peng Liangrui’s research group are the latest developed methods based on deep learning. The video text detection algorithm is based on a convolutional neural network, and the video text recognition algorithm is based on a recurrent neural network. The students who participated in the research project include Yan Ruijie, Xiang Donglai, Wang Yaqi, Wang Xuecheng, Chen Liren and Guo Jiaming, among others. The related research was supported by the 973 National Basic Research Program of China under Grant 2014CB340506 and National Natural Science Foundation of China under Grant U1636124.

Peng Liangrui has been engaged in multi-lingual OCR research for more than ten years. Peng Liangrui's research group belongs to the Laboratory for Intelligent Image and Document Information Processing at the Department of Electronic Engineering. The laboratory has developed world-leading innovative technologies in Chinese and multi-lingual OCR, multi-modal biometric identification, and video surveillance.

(Edited by Guo Lili, Zhu Lvhe)