Deep Template Matching for Pedestrian Attribute Recognition with the Auxiliary Supervision of Attribute-wise Keypoints

Abstract

Pedestrian Attribute Recognition (PAR) has aroused extensive attention due to its important role in video surveillance scenarios. In most cases, the existence of a particular attribute is strongly related to a partial region. Recent works design complicated modules, e.g., attention mechanism and proposal of body parts to localize the attribute corresponding region. These works further prove that localization of attribute specific regions precisely will help in improving performance. However, these part-information-based methods are still not accurate as well as increasing model complexity which makes it hard to deploy on realistic applications. In this paper, we propose a Deep Template Matching based method to capture body parts features with less computation. Further, we also proposed an auxiliary supervision method that use human pose keypoints to guide the learning toward discriminative local cues. Extensive experiments show that the proposed method outperforms and has lower computational complexity, compared with the state-of-the-art approaches on large-scale pedestrian attribute datasets, including PETA, PA-100K, RAP, and RAPv2zs.

Correlation Graph Convolutional Network for Pedestrian Attribute Recognition

“…To show the effectiveness of the proposed network, we compare it with several state-of-the-art PAR methods: HPNet [55], VeSPA [59], JRL [60], PGDM [30], MsVAA [37], GRL [31], VAC [15], JLPSL-PAA [61], RA [62], ALM [23], MT-CAS [38], DTM+AWK [63], JLAC [39], Baseline [32], SSC [64], DAFL [65], and VTB [47].…”

YinYang-Net: Complementing Face and Body Information for Wild Gender Recognition

“…Li et al [31] used pose data to aid in attribute body part localization, fusing features at multiple levels for attribute recognition. Zhang et al [32] proposed a Deep Template Matching method to capture body parts features, complemented by pose keypoints to guide discriminative cues learning. Liu et al [33] used body ROI and assigned attribute-specific weights based on extracted ROI proposals and attribute localization.…”

Reinforced Pedestrian Attribute Recognition with Group Optimization Reward

“…Depend on the class activation maps, Liu et al [32] proposed a Localization Guide Network (LGNet), which captures the activation box for each attribute by cropping the high response area of the corresponding activation map. DTM+AWK [33] leverages pose keypoints as auxiliary information to assist in positioning the attribute region. Tang et al [34] proposed an Attribute localization module (ALM) which can discover the most discriminative regions adaptively.…”

“…Tang et al [34] proposed an Attribute localization module (ALM) which can discover the most discriminative regions adaptively. The application of auxiliary information has been proven to be effective, however, they [31,32,33] largely depend on the accuracy of positioning.…”