IoU曲线上定位精度的核心所在：无标注推理时的边界优化
Where Grounding Accuracy Lives on the IoU Curve: Label-Free Inference-Time Boundary Refinement
浏览论文内容
中文总结 AI 辅助
本文提出无标注推理时的边界优化（LFPR）方法，在多个视觉-语言定位数据集上提升了IoU曲线各精度指标，证明指代选择与边界精度部分可分离，为定位精度优化提供了新思路。
中文摘要 AI 辅助
视觉-语言模型能够识别出正确的指代对象，但返回的边界框不够精确。我们研究在推理阶段，一个冻结的直接回答模型是否可以利用自身预测结果，在不访问目标标注的情况下分配一次额外的局部观测。无标注精度优化（LFPR）将预测的小区域送入更高分辨率的通道，在上下文裁剪区域内重新定位表达式，仅在固定几何约束下接受候选框，并返回固定的坐标中点。我们在三个证据层级上报告了结果：在31921个回顾性Ref-L4表达式上，LFPR将mAcc$_{0.5:0.95}$从72.947%提升至76.013%（Acc@0.5从88.531%提升至89.725%，Acc@0.9从55.788%提升至61.142%）；在30969个RefCOCO/RefCOCO+/RefCOCOg表达式上的冻结迁移，在Acc@0.5、mAcc和平均IoU上均提升了各数据集的性能（合并mAcc提升0.645，Acc@0.5提升0.817），而Acc@0.9整体不变：仅路由操作在此指标上提升1.162个点，但裁剪、约束和融合操作使其下降1.192，形成抵消而非无严格IoU效应；在图像不重叠的前瞻性Flickr30K Entities评估中，所有端点均得到提升（mAcc提升0.973，Acc@0.9提升1.022），单框变体下提升更显著（mAcc提升2.575，Acc@0.9提升3.689）；将相同操作应用于两个已发布的定位专家模型，在延迟约翻倍的情况下，所有端点均得到提升（EGM-4B/8B的Acc@0.9分别提升1.569/6.716），与专家训练结合而非替代它；移除相同候选框约束的无约束对照组在所有指标上均弱于现有模型，表明约束是关键组成部分。综合结果显示，指代对象选择与边界精度部分可分离，不同组件会在IoU曲线上移动相反区域，这是单一阈值无法揭示的行为。
英文摘要
Vision--language models can identify the correct referent while returning an imprecise bounding box. We study whether a frozen direct-answer model can use its own prediction to allocate one additional localized observation without accessing target annotations at inference. Label-free precision refinement (LFPR) routes predicted-small regions to a higher-resolution pass, re-grounds the expression inside a context crop, admits a candidate only under fixed geometric guards, and returns a fixed coordinate-wise midpoint. We report results across three evidence tiers. On 31,921 retrospective Ref-L4 expressions, LFPR raises mAcc$_{0.5:0.95}$ from 72.947\% to 76.013\% (Acc@0.5 88.531\%$\to$89.725\%, Acc@0.9 55.788\%$\to$61.142\%). A frozen transfer to 30,969 RefCOCO/RefCOCO+/RefCOCOg expressions improves every dataset at Acc@0.5, mAcc, and mean IoU (pooled mAcc $+0.645$, Acc@0.5 $+0.817$), while Acc@0.9 is unchanged overall: routing alone gains $+1.162$ points there, but crop, guards, and fusion give back $-1.192$, offsetting rather than showing no strict-IoU effect. A prospective, image-disjoint Flickr30K Entities evaluation improves every endpoint (mAcc $+0.973$, Acc@0.9 $+1.022$), more strongly under a single-box variant (mAcc $+2.575$, Acc@0.9 $+3.689$). The same operator applied to two released grounding specialists improves every endpoint (Acc@0.9 $+1.569$/$+6.716$ for EGM-4B/8B) at roughly twice the latency, composing with specialist training rather than replacing it. A genuine unguarded control (guard removed from the same candidates) underperforms the incumbent on every metric, showing the guard is load-bearing. Together, these results show that referent selection and boundary precision are partially separable, with different components moving opposing regions of the IoU curve -- behavior a single threshold cannot reveal.
发表机构
- Auckland University of Technology（奥克兰理工大学）
机构由 AI 辅助整理，请以论文原文为准。
