文献类型： 会议论文

第一作者： Yuecong Min

作者： Yuecong Min ¹ ; Peiqi Jiao ¹ ; Yanan Li ² ; Xiaotao Wang ² ; Lei Lei ² ; Xiujuan Chai ³ ; Xilin Chen ¹ ;

作者机构： 1.Key Lab of Intelligent Information Processing of Chinese Academy of Sciences (CAS), Institute of Computing Technology, CAS

2.Xiaomi Inc

3.Agricultural Information Institute, Chinese Academy of Agricultural Sciences

关键词： Deep feature embedding;Visaul sequence learning;Sign language recognition;Scene text recognition

会议名称： European Conference on Computer Vision

主办单位：

页码： 240-256

摘要： Connectionist Temporal Classification (CTC) is a popular objective function in sequence recognition, which provides supervision for unsegmented sequence data through aligning sequence and its corresponding labeling iteratively. The blank class of CTC plays a crucial role in the alignment process and is often considered responsible for the peaky behavior of CTC. In this study, we propose an objective function named RadialCTC that constrains sequence features on a hypersphere while retaining the iterative alignment mechanism of CTC. The learned features of each non-blank class are distributed on a radial arc from the center of the blank class, which provides a clear geometric interpretation and makes the alignment process more efficient. Besides, RadialCTC can control the peaky behavior by simply modifying the logit of the blank class. Experimental results of recognition and localization demonstrate the effectiveness of RadialCTC on two sequence recognition applications.

分类号： tp391.41-53