CASIA-E: A Large Comprehensive Dataset for Gait Recognition

• DOI: 10.1109/tpami.2022.3183288
• Publication Date: 2022-06-15T20:08:02Z
• Authors: Chunfeng Song; Yongzhen Huang; Weining Wang; Liang Wang
• Keywords: Deep Learning; Biometry; 0202 electrical engineering, electronic engineering, information engineering; Humans; Videotape Recording; Walking; 02 engineering and technology; Gait; Algorithms; Pattern Recognition, Automated

Gait recognition plays a special role in visual surveillance due to its unique advantage, <i>e.g.</i>, long-distance, cross-view and non-cooperative recognition. However, it has not yet been widely applied. One reason for this awkwardness is the lack of truly big dataset captured practical outdoor scenarios. Here, &#x201C;big&#x201D; at least means: (1) huge amount gait videos, (2) sufficient subjects, (3) rich attributes, (4) spatial temporal variations. Moreover, most existing large-scale datasets are collected indoors, which have few challenges from real scenes, such as dynamic complex background clutters, illumination variations, vertical view <i>etc</i>. In paper, we introduce newly built dataset, called CASIA-E. It contains more than one thousand people distributed over near million videos. Each person involves 26 angles varied appearances caused by changes bag carrying, dressing walking styles. The videos across five months three kinds scenes. Soft biometric features also recorded all subjects including age, gender, height, weight nationality. Besides, report an experimental benchmark examine some meaningful problems that well studied previously, influence million-level training angles, styles, thermal infrared modality. We believe will promote development both academic research industrial applications.