Real-Time Loop Closure Detection in Visual SLAM via NetVLAD and Faiss

TL;DR

This paper demonstrates that NetVLAD combined with Faiss enables real-time, accurate loop closure detection in visual SLAM, outperforming traditional bag-of-words methods like DBoW in robustness and speed.

Contribution

It empirically evaluates NetVLAD for LCD, introduces a new precision-recall metric, and shows real-time performance with Faiss acceleration as a practical alternative.

Findings

NetVLAD outperforms DBoW in accuracy and robustness.

Faiss enables real-time query speed for NetVLAD.

Proposed Fine-Grained Top-K metric better reflects LCD performance.

Abstract

Loop closure detection (LCD) is a core component of simultaneous localization and mapping (SLAM): it identifies revisited places and enables pose-graph constraints that correct accumulated drift. Classic bag-of-words approaches such as DBoW are efficient but often degrade under appearance change and perceptual aliasing. In parallel, deep learning-based visual place recognition (VPR) descriptors (e.g., NetVLAD and Transformer-based models) offer stronger robustness, but their computational cost is often viewed as a barrier to real-time SLAM. In this paper, we empirically evaluate NetVLAD as an LCD module and compare it against DBoW on the KITTI dataset. We introduce a Fine-Grained Top-K precision-recall curve that better reflects LCD settings where a query may have zero or multiple valid matches. With Faiss-accelerated nearestneighbor search, NetVLAD achieves real-time query speed while improving accuracy and robustness over DBoW, making it a practical drop-in alternative for LCD in SLAM.