Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Tran, Chi-Nguyen, Minh, Dao Sy Duy, Kiet, Huynh Trung, Quy, Nguyen Lam Phu, Pham, Phu-Hoa, Tran-Thanh, Long
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910231514054656
author Tran, Chi-Nguyen
Minh, Dao Sy Duy
Kiet, Huynh Trung
Quy, Nguyen Lam Phu
Pham, Phu-Hoa
Tran-Thanh, Long
author_facet Tran, Chi-Nguyen
Minh, Dao Sy Duy
Kiet, Huynh Trung
Quy, Nguyen Lam Phu
Pham, Phu-Hoa
Tran-Thanh, Long
contents Cross-view geo-localization (CVGL), which matches an oblique drone view to a geo-referenced satellite tile, has emerged as a key alternative for autonomous drone navigation when GNSS signals are jammed, spoofed, or unavailable. Despite strong recent progress, three limitations persist: (1) global-descriptor designs compress the patch grid into a single vector without separating layout from texture across the view gap; (2) altitude-related scale variation is retained in the learned embedding rather than marginalized; and (3) multi-objective training relies on hand-tuned scalars over losses on incompatible gradient scales. We propose SkyPart, a lightweight swappable head for patch-based vision transformers (ViTs) that institutes explicit part grouping over the patch grid. SkyPart has four theory-grounded components: (i) learnable prototypes competing for patch tokens via single-pass cosine assignment; (ii) altitude-conditioned linear modulation applied only during training, making the retrieval embedding altitude-free at inference; (iii) a graph-attention readout over active prototypes; and (iv) a Kendall uncertainty-weighted multi-objective loss whose stationary points are Pareto-stationary. At 26.95M parameters and 22.14 GFLOPs, SkyPart is the smallest among top-performing methods and sets a new state of the art on SUES-200, University-1652, and DenseUAV under a single-pass, no-re-ranking, no-TTA protocol. Its advantage over the strongest baseline widens under the ten-condition WeatherPrompt corruption benchmark.
format Preprint
id arxiv_https___arxiv_org_abs_2605_11654
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery
Tran, Chi-Nguyen
Minh, Dao Sy Duy
Kiet, Huynh Trung
Quy, Nguyen Lam Phu
Pham, Phu-Hoa
Tran-Thanh, Long
Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
Cross-view geo-localization (CVGL), which matches an oblique drone view to a geo-referenced satellite tile, has emerged as a key alternative for autonomous drone navigation when GNSS signals are jammed, spoofed, or unavailable. Despite strong recent progress, three limitations persist: (1) global-descriptor designs compress the patch grid into a single vector without separating layout from texture across the view gap; (2) altitude-related scale variation is retained in the learned embedding rather than marginalized; and (3) multi-objective training relies on hand-tuned scalars over losses on incompatible gradient scales. We propose SkyPart, a lightweight swappable head for patch-based vision transformers (ViTs) that institutes explicit part grouping over the patch grid. SkyPart has four theory-grounded components: (i) learnable prototypes competing for patch tokens via single-pass cosine assignment; (ii) altitude-conditioned linear modulation applied only during training, making the retrieval embedding altitude-free at inference; (iii) a graph-attention readout over active prototypes; and (iv) a Kendall uncertainty-weighted multi-objective loss whose stationary points are Pareto-stationary. At 26.95M parameters and 22.14 GFLOPs, SkyPart is the smallest among top-performing methods and sets a new state of the art on SUES-200, University-1652, and DenseUAV under a single-pass, no-re-ranking, no-TTA protocol. Its advantage over the strongest baseline widens under the ten-condition WeatherPrompt corruption benchmark.
title Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
url https://arxiv.org/abs/2605.11654