Not 3D Re-ID: Simple Single Stream 2D Convolution for Robust Video Re-identification

Abstract

Video-based person re-identification has received increasing attention recently, as it plays an important role within surveillance video analysis. Video-based Re-ID is an expansion of earlier image-based re-identification methods by learning features from a video via multiple image frames for each person. Most contemporary video Re-ID methods utilise complex CNNbased network architectures using 3D convolution or multibranch networks to extract spatial-temporal video features. By contrast, in this paper, we illustrate superior performance from a simple single stream 2D convolution network leveraging the ResNet50-IBN architecture to extract frame-level features followed by temporal attention for clip level features. These clip level features can be generalised to extract video level features by averaging without any significant additional cost. Our approach uses best video Re-ID practice and transfer learning between datasets to outperform existing state-of-the-art approaches on the MARS, PRID2011 and iLIDS-VID datasets with 89.62%, 97.75%, 97.33% rank-1 accuracy respectively and with 84.61% mAP for MARS, without reliance on complex and memory intensive 3D convolutions or multi-stream networks architectures as found in other contemporary work. Conversely, our work shows that global features extracted by the 2D convolution network are a sufficient representation for robust state of the art video Re-ID.

AA-RGTCN: reciprocal global temporal convolution network with adaptive alignment for video-based person re-identification

“…With the development of deep learning, many areas have gained great success including person re-identification. At first, most works extract features for each frame by CNN backbones such as ResNet-50 and use global average pooling or some simple weighted method to fuse them like Zhao et al ( 2019 ); Breckon and Alsehaim ( 2021 ) and Zheng et al ( 2016 ). It helps them to achieve significant results, but they merely use temporal clues to build up a discriminative representation.…”

“…The development of personal intelligence, such as home robotics technology, also needs the support of Re-ID technologies. Most relative works (Liu et al, 2015;Xu et al, 2017;Breckon and Alsehaim, 2021) can be summarized as two main steps: spatial feature extraction and temporal feature aggregation. CNN will be used as a spatial feature extractor for each frame, and the generated frame-level features will be sent to the aggregation part to produce a video-level feature.…”