Proposes weighted aggregation of clusters and self-distillation-driven token pruning to improve both accuracy and efficiency in ViT-based visual place recognition.
Towards seamless adaptation of pre-trained models for visual place recognition
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
fields
cs.CV 4years
2026 4representative citing papers
A geometry-aware second-stage network plus the XHZ ship dataset reduces mean VPR localization error by over 60% across multiple backbones in multi-floor maritime cabins.
Introduces LPQN and a two-phase perturbation method to degrade product quantization retrieval performance in visual localization pipelines, demonstrated via experiments in controlled and real-world settings.
RIA projects covariance descriptors from the SPD manifold into Euclidean space via Riemannian mappings to preserve structural invariants for VPR, matching supervised zero-shot performance and reaching SOTA with light fine-tuning.
citing papers explorer
-
Faster or Stronger: Towards Flexible Visual Place Recognition via Weighted Aggregation and Token Pruning
Proposes weighted aggregation of clusters and self-distillation-driven token pruning to improve both accuracy and efficiency in ViT-based visual place recognition.
-
From Open Waters to Enclosed Cabins: ProteusVPR for Cross-Scene Visual Place Recognition in Maritime Perception and Cabin Inspection
A geometry-aware second-stage network plus the XHZ ship dataset reduces mean VPR localization error by over 60% across multiple backbones in multi-floor maritime cabins.
-
Adversarial Attacks on Robot Localization Systems via Deep Feature Perturbation
Introduces LPQN and a two-phase perturbation method to degrade product quantization retrieval performance in visual localization pipelines, demonstrated via experiments in controlled and real-world settings.
-
Beyond First-Order: Learning Riemannian Geometries for Invariant Visual Place Recognition
RIA projects covariance descriptors from the SPD manifold into Euclidean space via Riemannian mappings to preserve structural invariants for VPR, matching supervised zero-shot performance and reaching SOTA with light fine-tuning.