PARC-Loc: Text-to-Point-Cloud Localization with Partial Assignment and Relational Consistency

2026-10-07

AI Summary

This study addresses layout ambiguity caused by repetitive objects and missing boundary evidence in subgraphs for text-to-point cloud localization. To tackle these challenges, we propose a coarse-to-fine cross-modal localization framework that jointly models object compatibility and spatial relationships to optimize subgraph selection. Furthermore, partial assignment and relation consistency mechanisms are introduced to accommodate unmatched elements while expanding neighborhood context for enhanced evidence completeness. Precise localization is achieved by integrating candidate-level evaluation, neural similarity fusion, and PARC-based cross-modal attention guidance. Experimental results on the KITTI360Pose dataset demonstrate that our method improves Top-1 recall by 34%, significantly outperforming existing baselines.

Abstract

Text-to-point-cloud localization estimates a position in a city-scale 3D map from descriptions of surrounding objects. Existing coarse-to-fine methods retrieve submaps using aggregate learned compatibility and then localize within a selected submap. However, repetitive or similar urban objects can inflate the embedding similarity between the query and multiple submaps, even when the instance layout within a submap violates the query description. Meanwhile, query-relevant instances often span submap boundaries, leaving the retrieved submap with incomplete contextual evidence. We term these failure modes layout-inconsistent aliasing and boundary evidence incompleteness, respectively. To address them, we propose PARC-Loc, a coarse-to-fine localization framework built on Partial Assignment with Relational Consistency (PARC). PARC jointly models hint-object compatibility and pairwise spatial relations, allowing unmatched elements while favoring assignments consistent with the queried layout. At the coarse stage, its candidate-level assessment complements neural similarity for layout-consistent submap selection. At the fine stage, the context is expanded with query-relevant instances from adjacent submaps, while PARC yields object-level matching weights that guide cross-modal attention. Extensive experiments on KITTI360Pose and CityLoc show that PARC-Loc outperforms conventional coarse-to-fine baselines. On KITTI360Pose, our method improves Top-1 localization recall at 5 m from 0.50 to 0.67, achieving a 34% relative gain over the strongest baseline.

Problem

• text-to-point-cloud localization
• layout-inconsistent aliasing
• boundary evidence incompleteness
• coarse-to-fine localization

Innovation

• Text-to-Point-Cloud Localization
• Partial Assignment
• Relational Consistency
• Coarse-to-Fine Framework
• Cross-modal Attention