Image-to-Video Re-Identification via Mutual Discriminative Knowledge Transfer

TL;DR

This paper introduces a mutual discriminative knowledge distillation framework with a novel triplet contrast loss to improve image-to-video re-identification by effectively transferring richer video representations to image-based models.

Contribution

It proposes a new mutual discriminative knowledge distillation method with triplet contrast loss for better transfer of video features to image models in re-identification tasks.

Findings

Outperforms existing methods on MARS, DukeMTMC-VideoReID, VeRi-776 datasets.

Triplet contrast loss effectively transfers local discriminative features.

Mutual learning regularizes training for improved performance.

Abstract

The gap in representations between image and video makes Image-to-Video Re-identification (I2V Re-ID) challenging, and recent works formulate this problem as a knowledge distillation (KD) process. In this paper, we propose a mutual discriminative knowledge distillation framework to transfer a video-based richer representation to an image based representation more effectively. Specifically, we propose the triplet contrast loss (TCL), a novel loss designed for KD. During the KD process, the TCL loss transfers the local structure, exploits the higher order information, and mitigates the misalignment of the heterogeneous output of teacher and student networks. Compared with other losses for KD, the proposed TCL loss selectively transfers the local discriminative features from teacher to student, making it effective in the ReID. Besides the TCL loss, we adopt mutual learning to regularize both the teacher and student networks training. Extensive experiments demonstrate the effectiveness of our method on the MARS, DukeMTMC-VideoReID and VeRi-776 benchmarks.