StructVPR: Distill Structural Knowledge with Weighting Samples for Visual Place Recognition

TL;DR

StructVPR introduces a novel training method that leverages structural knowledge from segmentation images and weighted distillation to improve global features for visual place recognition, outperforming existing methods.

Contribution

The paper proposes a new training architecture that enhances global feature stability in VPR by using segmentation-based structural knowledge and weighted knowledge distillation, avoiding online segmentation during testing.

Findings

Achieves state-of-the-art performance on multiple benchmarks.

Outperforms many two-stage approaches with only global retrieval.

Maintains low computational cost with added re-ranking.

Abstract

Visual place recognition (VPR) is usually considered as a specific image retrieval problem. Limited by existing training frameworks, most deep learning-based works cannot extract sufficiently stable global features from RGB images and rely on a time-consuming re-ranking step to exploit spatial structural information for better performance. In this paper, we propose StructVPR, a novel training architecture for VPR, to enhance structural knowledge in RGB global features and thus improve feature stability in a constantly changing environment. Specifically, StructVPR uses segmentation images as a more definitive source of structural knowledge input into a CNN network and applies knowledge distillation to avoid online segmentation and inference of seg-branch in testing. Considering that not all samples contain high-quality and helpful knowledge, and some even hurt the performance of distillation, we partition samples and weigh each sample's distillation loss to enhance the expected knowledge precisely. Finally, StructVPR achieves impressive performance on several benchmarks using only global retrieval and even outperforms many two-stage approaches by a large margin. After adding additional re-ranking, ours achieves state-of-the-art performance while maintaining a low computational cost.