OASIS-Map: Object-Level Change Detection in Multi-Session Mapping using Semantic Correspondence Matching

2026-07-16

AI Summary

This work addresses the challenge of rapidly outdated static maps in multi-session mapping scenarios, where objects frequently appear, disappear, move, or are replaced—particularly under partial observability, occlusion, and imperfect segmentation that hinder reliable cross-session object association. To tackle this, the authors propose a dense patch-level semantic correspondence-based method for object-level change detection and incremental association. By establishing semantic correspondences between observations from revisited sessions, the approach enables temporally and spatially consistent object-level map updates and scene change detection. The method achieves an F1 score of 0.783 in a parking lot vehicle replacement scenario and 0.667 on the 3RScan dataset for object movement association, demonstrating robustness and effectiveness in complex real-world environments.

Abstract

Map representations which are consistent across repeated visits to a real-world semi-static environment are very useful for long-term robotic inspection. In such settings, the scene may evolve while the robot is absent, with objects appearing, disappearing, moving, or being replaced, quickly making a static map outdated. Existing change-detection methods reason through geometry, category-level semantics, or object persistence. However, achieving reliable object association across revisits remains a key challenge, especially under partial views, occlusion, and imperfect segmentation. In this work, we propose OASIS-Map, a multi-session mapping system that maintains a spatio-temporally consistent object-level map by establishing dense patch-level semantic correspondences between temporal observations. These correspondences detect where the scene has changed and incrementally associate objects across revisits as the robot re-observes the environment. We demonstrate OASIS-Map on three challenging real-world scenarios: object rearrangements in 3RScan, visually similar car replacements in a car park, and large-scale scene changes in an outdoor market. We achieve 0.783 F1 on change detection in a car replacement scenario in a car park and 0.667 F1 on moved object association in 3RScan. https://dynamic.robots.ox.ac.uk/projects/oasis-map/

Problem

• change detection
• object association
• multi-session mapping
• semantic correspondence
• spatio-temporal consistency

Innovation

• semantic correspondence
• object-level change detection
• multi-session mapping
• spatio-temporal consistency
• dense patch matching