vMAP: Vectorised Object Mapping for Neural Field SLAM

TL;DR

vMAP is a neural field SLAM system that efficiently models multiple objects in real-time without requiring 3D priors, significantly improving scene and object reconstruction quality.

Contribution

It introduces a vectorised training approach enabling real-time, object-level neural field mapping with no need for 3D priors, handling up to 50 objects simultaneously.

Findings

Real-time map update at 5Hz for multiple objects

Enhanced scene and object reconstruction quality

Efficient neural object modelling without 3D priors

Abstract

We present vMAP, an object-level dense SLAM system using neural field representations. Each object is represented by a small MLP, enabling efficient, watertight object modelling without the need for 3D priors. As an RGB-D camera browses a scene with no prior information, vMAP detects object instances on-the-fly, and dynamically adds them to its map. Specifically, thanks to the power of vectorised training, vMAP can optimise as many as 50 individual objects in a single scene, with an extremely efficient training speed of 5Hz map update. We experimentally demonstrate significantly improved scene-level and object-level reconstruction quality compared to prior neural field SLAM systems. Project page: https://kxhit.github.io/vMAP.