Semi-supervised Vector-Quantization in Visual SLAM using HGCN

TL;DR

This paper introduces semi-supervised loop closure detection methods using Hyperbolic Graph Convolutional Networks in visual SLAM, demonstrating improved efficiency and accuracy over traditional unsupervised approaches.

Contribution

It presents novel semi-supervised HGCN-based techniques for loop closure detection and extends ORB-SLAM with these methods, showing advantages in scalability and performance.

Findings

HGCN-FABMAP requires more cluster centroids for effective detection.

HGCN-ORB is more memory-efficient than HGCN-FABMAP.

HGCN-based methods outperform traditional algorithms in accuracy and scalability.

Abstract

In this paper, two semi-supervised appearance based loop closure detection technique, HGCN-FABMAP and HGCN-BoW are introduced. Furthermore an extension to the current state of the art localization SLAM algorithm, ORB-SLAM, is presented. The proposed HGCN-FABMAP method is implemented in an off-line manner incorporating Bayesian probabilistic schema for loop detection decision making. Specifically, we let a Hyperbolic Graph Convolutional Neural Network (HGCN) to operate over the SURF features graph space, and perform vector quantization part of the SLAM procedure. This part previously was performed in an unsupervised manner using algorithms like HKmeans, kmeans++,..etc. The main Advantage of using HGCN, is that it scales linearly in number of graph edges. Experimental results shows that HGCN-FABMAP algorithm needs far more cluster centroids than HGCN-ORB, otherwise it fails to detect loop closures. Therefore we consider HGCN-ORB to be more efficient in terms of memory consumption, also we conclude the superiority of HGCN-BoW and HGCN-FABMAP with respect to other algorithms.