Distillation Improves Visual Place Recognition for Low Quality Images

TL;DR

This paper introduces a knowledge-distillation approach to enhance visual place recognition accuracy in low-quality images by learning discriminative features from high-quality images, validated across multiple datasets and modalities.

Contribution

It proposes a novel distillation method with specific loss functions to improve descriptor quality for low-quality images in VPR, including video data.

Findings

Significant improvements in VPR recall rates under JPEG compression, resolution reduction, and video quantization.

Effective application of the method across multiple VPR algorithms and datasets.

Fills a research gap by analyzing video-based data impact on VPR performance.

Abstract

Real-time visual localization often utilizes online computing, for which query images or videos are transmitted to remote servers for visual place recognition (VPR). However, limited network bandwidth necessitates image-quality reduction and thus the degradation of global image descriptors, reducing VPR accuracy. We address this issue at the descriptor extraction level with a knowledge-distillation methodology that learns feature representations from high-quality images to extract more discriminative descriptors from low-quality images. Our approach includes the Inter-channel Correlation Knowledge Distillation (ICKD) loss, Mean Squared Error (MSE) loss, and Triplet loss. We validate the proposed losses on multiple VPR methods and datasets subjected to JPEG compression, resolution reduction, and video quantization. We obtain significant improvements in VPR recall rates under all three tested modalities of lowered image quality. Furthermore, we fill a gap in VPR literature on video-based data and its influence on VPR performance. This work contributes to more reliable place recognition in resource-constrained environments.