跳到正文
HuggingFace Daily Papers·· 2 天前

OBB-PBD:将并行框解码扩展到航空图像指向性表达定位的有向边界框

OBB-PBD: Extending Parallel Box Decoding to Oriented Bounding Boxes for Referring Expression Grounding in Aerial Imagery

AI 导读

OBB-PBD 将原本只支持四坐标水平框的并行框解码(PBD)扩展为六槽有向表示,采用双角度参数化。在 AVVG 基准全部 6139 条测试记录上,rotIoU 从零样本基线的 0.1253 提升到 0.303±0.012;相比对齐包络框的匹配对照,学习有向框额外带来 0.055±0.008 rotIoU,而仅做块扩展的增益与零无显著差异。

正文 · 原文

View PDF HTML (experimental)

Abstract:Referring expression grounding in overhead imagery requires oriented bounding boxes rather than horizontal ones, since aerial objects appear at arbitrary orientations. Most grounding models emit box coordinates sequentially; Parallel Box Decoding (PBD) instead predicts all geometry slots in one forward pass, but only for four-coordinate horizontal boxes. We extend this block-aligned interface to a six-slot oriented representation using a double-angle parameterization. Rather than assuming correctness from a decreasing loss curve, we verify the extension with image/prompt-shuffle controls and visibility/leakage probes, which show that the corrected model depends on both the image and the prompt rather than on prompt statistics alone. On the aerial vehicle visual grounding (AVVG) benchmark, whose targets are genuinely oriented, the extended model improves rotated intersection over union (rotIoU) from 0.1253 for the zero-shot baseline to 0.303 +/- 0.012 (mean +/- standard deviation over three training seeds) across the complete 6139-record test split spanning all 120 images. Learning oriented boxes rather than their axis-aligned envelopes adds 0.055 +/- 0.008 rotIoU beyond a matched control (per-seed 95% confidence intervals all excluding zero), whereas block expansion alone yields no rotIoU gain distinguishable from zero.
Comments: Published in Remote Sensing 18(20), 3535 (2026). Code and checkpoints: this https URL and this https URL
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Cite as: arXiv:2610.35450 [cs.CV]
  (or arXiv:2610.35450v1 [cs.CV] for this version)
  https://doi.org/10.48550/arXiv.2610.35450

arXiv-issued DOI via DataCite (pending registration)

Journal reference: Remote Sensing 18(20), 3535 (2026)
Related DOI: https://doi.org/10.3390/rs18203535

DOI(s) linking to related resources

Submission history

From: Changwoo Song [view email]
[v1] Mon, 26 Oct 2026 13:48:32 UTC (2,343 KB)

来源:HuggingFace Daily Papers · arxiv.org