Introspection in learned semantic scene graph localisation

Conference item

This work investigates how semantics influence localisation performance and robustness in a learned selfsupervised, contrastive semantic localisation framework. After training a localisation network on both original and perturbed maps, we conduct a thorough post-hoc introspection analysis to probe whether the model filters environmental noise and prioritises distinctive landmarks over routine clutter. We validate various interpretability methods and present a comparative reliability analysis. Integrated gradients and Attention Weights consistently emerge as the most reliable probes of learned behaviour. A semantic class ablation further reveals an implicit weighting in which frequent objects are often down-weighted. Overall, the results indicate that the model learns noise-robust, semantically salient relations about place definition, thereby enabling explainable registration under challenging visual and structural variations.

Authors: Charvi Bissessur, M; Panagiotaki, E; De Martini, D

• Publication status: Accepted
• Peer review status: Peer reviewed
• Publisher: IEEE
• Acceptance date: 2025-10-04
• Event title: IROS 2025 FAST workshop
• Event location: Hangzhou, China
• Event website: https://sites.google.com/view/iros2025fastworkshop
• Event start date: 2025-10-24
• Event end date: 2025-10-24
• Language: English