Learning to Segment Dynamic Objects using SLAM Outliers

TL;DR

This paper introduces a method that automatically learns to segment dynamic objects using SLAM outliers, enabling improved SLAM robustness by removing features on dynamic objects during operation.

Contribution

It presents a novel approach that uses SLAM outliers for dynamic object segmentation, requiring only one training sequence per object, and integrates this into existing SLAM systems.

Findings

Outperforms state-of-the-art on TUM RGB-D dataset in monocular mode.

Introduces a new stereo dataset with challenging consensus inversion scenarios.

Achieves better SLAM robustness by removing dynamic object features at runtime.

Abstract

We present a method to automatically learn to segment dynamic objects using SLAM outliers. It requires only one monocular sequence per dynamic object for training and consists in localizing dynamic objects using SLAM outliers, creating their masks, and using these masks to train a semantic segmentation network. We integrate the trained network in ORB-SLAM 2 and LDSO. At runtime we remove features on dynamic objects, making the SLAM unaffected by them. We also propose a new stereo dataset and new metrics to evaluate SLAM robustness. Our dataset includes consensus inversions, i.e., situations where the SLAM uses more features on dynamic objects that on the static background. Consensus inversions are challenging for SLAM as they may cause major SLAM failures. Our approach performs better than the State-of-the-Art on the TUM RGB-D dataset in monocular mode and on our dataset in both monocular and stereo modes.