GrabVG: Graph-Attentive Binding for Visual Grounding in UAV Imagery

Chaowei Wang; Yan Di; Jingjun Sun; Baozhe Liu; Jiaxu Tian; Yuheng Li; Guangqian Guo; Shan Gao

cs.CVAug 19, 2026

GrabVG: Graph-Attentive Binding for Visual Grounding in UAV Imagery

Authors: Chaowei Wang, Yan Di, Jingjun Sun, Baozhe Liu, Jiaxu Tian, Yuheng Li, Guangqian Guo, Shan Gao

Organizations: Northwestern Polytechnical University · Harbin Institute of Technology · The Hong Kong Polytechnic University

Abstract

Visual grounding in Unmanned Aerial Vehicle (UAV) imagery aims to localize a target object in complex bird's-eye-view scenes according to a natural language description. However, the abundance of small, densely distributed, and visually similar objects creates high visual redundancy, while repetitive local configurations give rise to strong topological ambiguity. Existing approaches mainly focus on visual--language feature alignment or dense contextual interaction, yet they struggle to distinguish subtle inter-instance differences and effectively exploit spatial topological structures, leading to inaccurate grounding in highly crowded scenarios. To address these challenges, we propose $\textbf{GrabVG}$ , a novel visual grounding framework inspired by human visual search. GrabVG explicitly decomposes grounding into two sequential stages: $\textit{preattentive hypothesis search}$ and $\textit{graph-attentive feature binding}$ . Specifically, we first generate a compact set of reliable object hypotheses through distillation-guided proposal induction and text-aware hypothesis filtering, substantially reducing background distractions and semantic mismatches. These hypotheses are then organized into a sparse graph, where language-guided intra-instance visual cues and inter-instance topological relationships are jointly bound and propagated via graph attention, enabling efficient spatial reasoning and accurate target localization. Extensive experiments on AerialVG and AerialSense show that GrabVG achieves a favorable accuracy--speed trade-off, reaching 67.31 $\%$ and 80.34 $\%$ [email protected] and outperforming the corresponding baselines by 10.55 and 8.76 percentage points, respectively.

Figures & tables

Figure 1: Comparison of visual grounding paradigms. Baseline [ 19 ] : Existing methods rely on global decoding or dense relational modeling, often leaving visually similar hypotheses insufficiently distinguished. Human Visual Search [ 30 , 25 ] : Preattentive guidance first narrows the search, followed by focused binding of discriminative evidence. Ours: GrabVG follows this progressive process through hypothesis induction, hypothesis filtering, and graph-attentive feature binding.

Figure 2: Overall architecture of GrabVG. (a) Given a UAV image and a referring expression, Distillation-Guided Proposal Induction first produces an initial hypothesis set $\mathcal{H}^{0}$ . Text-Aware Hypothesis Filtering then removes background queries and retains an expression-aligned candidate subset. The resulting hypotheses $\mathcal{H}^{f}$ are processed by Intra-Instance Appearance Binding and Inter-Instance Topological Binding, followed by CRS for final target selection. (b) Intra-Instance Appearance Binding samples language-guided visual features from each hypothesis region and its immediate surroundings to obtain appearance-aware representations $\mathbf{Q}^{a}$ . (c) Inter-Instance Topological Binding exchanges relational information over a sparse geometric graph using language-conditioned edge features, producing graph-aware representations $\mathbf{Q}^{g}$ .

Method	Top1 (%)	Top5 (%)	mIoU (%)
Zero-Shot Methods
ReX-Omni [ 12 ]	28.37	30.09	-
Supervised Methods
TransVG [ 7 ]	11.53	13.68	-
D-MDETR [ 22 ]	19.87	29.87	-
G-DINO [ 20 ]	29.36	78.87	-

Table 1: Comparison with state-of-the-art methods on the AerialVG test set. “*” indicates results obtained with hyperparameter optimization. Best results are in bold . Top1 and Top5 denote [email protected] and Top-5 [email protected], respectively.

Large Multimodal Models (Zero-Shot)
Method	Type	Top1 (%)	mIoU (%)
ReX-Omni [ 12 ]	Zero-shot	71.47	-
Domain-Specific Models
AerialVG [ 19 ]	Supervised	71.58	-
PropVG* [ 5 ]	Supervised	61.30	67.76
GrabVG (Ours)	Supervised	80.34	72.00

Table 2: Performance on the AerialSense benchmark. We compare GrabVG with a recent zero-shot multimodal model and aerial-specific supervised baselines.

Baseline	GrabVG Components				Test		Val
Baseline	$\boldsymbol{\mathcal{Y}_{tea}}$	$\boldsymbol{\mathcal{F}_{app}}$	$\boldsymbol{\mathcal{F}_{topo}}$	$\boldsymbol{\mathcal{F}_{filt}}$	Top1	Top5	Top1	Top5
✓					56.76	74.05	55.96	72.28
✓	✓				62.84	85.54	61.82	85.04
✓	✓	✓			65.59	87.89	65.42	87.13
✓	✓		✓		65.10	88.02	64.15	87.86
✓	✓	✓	✓		67.01	89.52	67.13	88.96

Table 3: Component-wise Ablation on AerialVG. The baseline is a hyperparameter-optimized PropVG model trained with the original ground-truth annotations.

\mathcal{Y}_{tea}

denotes fixed expression-aligned pseudo annotations generated offline by a trained AerialVG teacher;

\mathcal{F}_{app}

\mathcal{F}_{topo}

, and

\mathcal{F}_{filt}

denote appearance binding, topological binding, and text-aware filtering, respectively.

\mathcal{F}_{topo}

uses a symmetric

k

-NN graph.

Pipeline Strategy	Top1 (%)	Top5 (%)
$\mathcal{F}_{ind}+\mathcal{F}_{topo}$ (no appearance binding)	65.10	88.02
$\mathcal{F}_{ind}+\mathcal{F}_{app}+\mathcal{F}_{topo}+\mathrm{NMS}$	67.25	90.03
$\mathcal{F}_{ind}+\mathcal{F}_{filt}+\mathcal{F}_{app}+\mathcal{F}_{filt}+\mathcal{F}_{topo}$	66.75	89.46
$\mathcal{F}_{ind}+\mathcal{F}_{filt}+\mathcal{F}_{app}+\mathcal{F}_{topo}$	67.31	89.43

Table 4: Pipeline design choices on AerialVG. Only the post-induction candidate-processing strategy is changed.

Anchor Topology	Scales	Top1 (%)	Top5 (%)
Single Ring	$[1.0]$	64.59	87.15
Double Ring	$[0.5,1.0]$	64.57	87.51
Double Ring	$[1.0,1.2]$	64.36	87.56
Triple Ring (Ours)	$[0.5,1.0,1.2]$	65.59	87.89

Table 5: Ablation of Intra-Instance Appearance Binding on AerialVG. All variants use proposal induction followed by appearance binding; filtering and topological binding are disabled.

Topology Strategy
Relational Topology	Layers	Top1 (%)	Top5 (%)
Dense (AerialVG-style)	4	63.98	87.07
KNN ( $k=6$ )	4	67.31	89.43
Delaunay Triangulation	4	66.21	88.65
Graph Depth with KNN ( $k=6$ )
KNN	2	64.34	88.32

Table 6: Ablation of Inter-Instance Topological Binding on AerialVG. The upper block changes the graph structure; the lower block changes only the depth of the default KNN graph.

Prior Source	Source Baseline	With GrabVG	Gain (pp)
Training GT Boxes	56.76	58.94	+2.18
LLM + Grounding-DINO [ 20 ]	56.83	60.06	+3.23
LLM + SAM3 [ 1 ]	56.97	60.36	+3.39
AerialVG-Derived [ 19 ]	62.84	67.31	+4.47
Oracle (All GT Boxes) ∗	69.62	74.23	+4.61

Table 7: Robustness to hypothesis priors on the AerialVG test set. All values are Top1 ([email protected]). LLM denotes Qwen3 [ 32 ] . ∗ The oracle setting supplies all ground-truth boxes and omits

\mathcal{F}_{filt}

Existing Baselines
Method	FPS (img/s)
AerialVG [ 19 ]	13.65
OTA-Det [ 29 ]	32.00
GrabVG Variants
$\mathcal{F}_{ind}$	30.32
+ $\mathcal{F}_{app}$ (Intra-binding)	30.03

Table 8: Inference-speed comparison. FPS results for GrabVG, its component variants, and existing baselines under the same evaluation setup. All FPS results are measured on a single NVIDIA RTX 4090 GPU.

Figure 3: Qualitative ablation of progressive feature binding. Appearance binding corrects the target attribute, while topological binding resolves the remaining spatial ambiguity.

Figure 4: Visualization of Graph-Attentive Feature Binding. Top: sparse graph edges and directional attention distributions, where thick red arrows indicate the highest-attended edges. Bottom: top- $8$ adaptive sampling points highlighting appearance-relevant regions.

Visual Grounding

Visual grounding connects language to the image regions, objects, or pixels that support a description, question, or action.

See all 155 papers

VL-UniTrack: A Unified Framework with Visual-Language Prompts for UAV-Ground Visual Tracking
May 6, 2026Boyue Xu, Ruichao Hou, Tongwei Ren +1Visual Object Tracking Multi-View Learning
State Key Laboratory for Novel Software Technology, Nanjing University
Training-Free Open-Vocabulary Visual Grounding for Remote Sensing Images and Videos
Jun 15, 2026Ke Li, Di Wang, Yongshan Zhu +5Remote Sensing Image Understanding Temporal Video Grounding
School of Computer Science and Technology, Xidian University, Xi’an 710071, China · Interdisciplinary Institute of Artificial Intelligence, Xidian University, Xi’an, Shaanxi 710126, China · School of Artificial Intelligence, Xidian University, Xi’an 710071, China +2
UniGround: Universal 3D Visual Grounding via Training-Free Scene Parsing
Mar 9, 2026Jiaxi Zhang, Yunheng Wang, Wei Lu +83D Visual Grounding 3D VLMs
The Hong Kong University of Science and Technology (Guangzhou) · JD Explore Academy · Zhejiang University

May 6, 2026J/K move · Enter open · S save

VL-UniTrack: A Unified Framework with Visual-Language Prompts for UAV-Ground Visual Tracking

Boyue Xu, Ruichao Hou, Tongwei Ren, Gangshan Wu

State Key Laboratory for Novel Software Technology, Nanjing University

Visual Object Tracking Multi-View Learning Visual Prompt Learning

Open paper AskPDF arXiv

UAV-ground visual tracking (UGVT) aims to simultaneously track the same object from both the UAV and the ground view. However, existing two-stream methods suffer from isolated feature extraction and rely heavily on implicit appearance matching, which struggles to establish reliable correspondence under drastic view differences, leading to tracking unreliability. To address these limitations, we propose VL-UniTrack, a fully unified framework enhanced by visual-language prompts. By encoding features from both views within a single shared encoder, our method breaks the barrier of feature isolation to facilitate sufficient cross-view interaction. To overcome the ambiguity caused by relying solely on appearance matching, we design visual-language geometric prompting module, which fuses language descriptions with visual features to generate learnable prompts. These prompts are then fed into our prompt-guided cross-view adapter module to enable sufficient cross-view feature interaction and to guide the learning of view-specific feature representations. Furthermore, a confidence-modulated mutual distillation loss is proposed to regularize the training by mitigating noise propagation. Extensive experiments demonstrate that our method achieves state-of-the-art performance on the latest benchmark. The code can be downloaded in https://github.com/xuboyue1999/VL-UniTrack.git

Abstract

Visual Grounding

Explore similar work

VL-UniTrack: A Unified Framework with Visual-Language Prompts for UAV-Ground Visual Tracking

Training-Free Open-Vocabulary Visual Grounding for Remote Sensing Images and Videos

UniGround: Universal 3D Visual Grounding via Training-Free Scene Parsing