Pith. sign in

REVIEW 3 major objections 5 minor 65 references

Advancing MLLM-based UAV Image Understanding and Reasoning: A Benchmark and a Training-Free Multi-Agent System

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A training-free multi-agent system with a 32B open-source model reaches 77.0% accuracy on a new UAV question-answer benchmark, beating Gemini 3 Pro by 4.0%.

desk verdict A serious benchmark-plus-agent paper whose headline claim is plausible but can't be fully trusted until the benchmark and code are released. read the letter →

arxiv 2608.11738 v1 pith:ILHECK6T submitted 2026-08-12 cs.CV cs.AI

Haoyu Zhang , Shuoxun Zhang , Peng Ye , Lin Zhang , Jiakang Yuan , Shenghong Yi , Yuening Wang , Tao Chen This is my paper · ORCID
classification cs.CVcs.AI
keywords UAVaerialimageunderstandingmultimodallargelanguagemodelsmulti-agentsystemsbenchmarkvisualgroundingreasoningtraining-freetooluse
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that the gap between open-source and proprietary multimodal models on drone-view image understanding can be closed without any training, by reorganizing how a model uses tools and verifies its own reasoning. To do that, it builds UAVQA-Bench, a 1,500-question benchmark with human-annotated answers and bounding boxes spanning 16 tasks over 13 public UAV datasets, and reports that its training-free multi-agent system UAV-MAS reaches 77.0% overall accuracy with a 32B open-source model, 4.0 points above Gemini 3 Pro. It also reports the 8B version gains 8.7 points over its base model. If the benchmark is representative, the result would mean domain-aware orchestration, not larger models or fine-tuning, can be the deciding factor for aerial intelligence.

What carries the argument

The central object is UAV-MAS, a training-free multi-agent loop built on three interacting mechanisms. DSPE routes each query to a small set of aerial-specific tools through per-tool activation agents; CAIR wraps the ReAct loop, which interleaves reasoning and tool actions, with a perceptual verification agent that decides whether new evidence updates or preserves the current answer; DAAS scores each reasoning step, prunes a branch only when two consecutive scores fall below a difficulty-derived threshold, and selects the best path by average score. Together they convert a single MLLM call into a structured, self-verifying tool-use process, which is what the paper credits for the accuracy gains.

What would settle it

Have a different annotation team independently re-sample images from the same 13 public UAV datasets, build a held-out question set with the same 16 task templates, and run UAV-MAS-32B and Gemini 3 Pro on it; if the 4.0-point margin disappears or reverses, the original benchmark's sample selection rather than system capability produced the reported lead.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that UAV aerial understanding fails in three identifiable ways: tools trained on ground-level images do not match aerial patterns, errors made by tools cascade through multi-step chains, and fixed linear reasoning does not adapt to query difficulty. After evaluating open- and closed-source MLLMs and agent systems on UAVQA-Bench, the paper proposes UAV-MAS, in which a Domain-Specific Perception Engine provides aerial-adapted tools such as zooming, description, depth estimation, semantic grounding, and de-hallucinated open-vocabulary detection; a Context-Aware Iterative Refinement module verifies each tool step and updates the accumulated answer state only when consistent; and a Difficulty-Aware Adaptive Search prunes low-confidence branches only after two consecutive low scores while selecting the final path by average node score. The reported result is 77.0% overall accuracy for the 32B configuration, surpassing Gemini 3 Pro by 4.0%, and 70.47% for the 8B configuration, an 8.7-point gain over its base model; the system also holds up on the independent CHOICE benchmark at 75.23%.

Load-bearing premise

UAVQA-Bench's 1,500 manually selected images and questions fairly represent UAV aerial understanding, so the 4.0-point lead over Gemini 3 Pro reflects real capability rather than benchmark selection.

Editorial extensions

If this is right

  • A 32B open-source model can surpass a top closed-source model on a broad UAV QA benchmark when equipped with domain tools and verification, without fine-tuning.
  • An 8B agent is enough to lift accuracy by 8.7 points on the same benchmark, suggesting small models benefit most from structured tool use.
  • Difficulty-adaptive search beats brute-force repeated sampling: at similar accuracy, UAV-MAS uses 57.7% less time and fewer model and tool calls than majority vote at 3.
  • The benchmark provides a single yardstick for comparing perception, counting, spatial reasoning, and grounding in overhead imagery.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The three diagnosed failure modes are likely to appear in adjacent overhead-imagery settings such as satellite imagery, traffic cameras, and agricultural surveys, so the DSPE-CAIR-DAAS recipe may transfer to those domains with a different toolset.
  • Because the benchmark was built by the same group that designed the system, an independent re-annotation or fresh-image sample is the decisive test of whether the 4.0-point margin reflects capability rather than benchmark selection.
  • If the benchmark's closed-ended format were replaced with free-form answers, rankings could shift, since open-ended responses reward different reasoning behaviors than option selection.
  • The per-tool activation agents and score agents could themselves be replaced by a learned router or a smaller distilled model, which would shrink latency while preserving most of the gain.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper introduces UAVQA-Bench, a 1,500-sample human-annotated benchmark for UAV aerial image understanding and reasoning covering 6 capability dimensions and 16 tasks in multiple-choice and visual grounding formats, drawn from 13 public UAV datasets. The authors evaluate a range of open- and closed-source MLLMs and agent-based systems, identify three failure modes (domain-toolset mismatch, error propagation, static reasoning), and propose UAV-MAS, a training-free multi-agent system with a Domain-Specific Perception Engine (DSPE), a Context-Aware Iterative Refinement module (CAIR), and a Difficulty-Aware Adaptive Search mechanism (DAAS). The central claim is that UAV-MAS with a 32B open-source MLLM achieves 77.0% overall accuracy on UAVQA-Bench, surpassing Gemini 3 Pro by 4.0%, and that the 8B variant improves by 8.7% over its base model. A cross-dataset experiment on the CHOICE benchmark reports generalization gains for the 8B system using only generic zooming and description tools.

Significance. If the central claim holds, the paper makes a meaningful contribution: it provides a broader UAV question-answering benchmark than existing options, and it demonstrates a training-free agent architecture that can make an open-source 32B MLLM competitive with a proprietary frontier model on a diverse aerial understanding suite. The proposed UAV-MAS is clearly described and modular, and the module-level and within-module ablations in Tables IV and V are a useful first analysis of where the gains come from. The cross-dataset CHOICE result (Table VII) is genuine independent evidence for the reasoning components, and the efficiency comparison in Table VI addresses an important practical concern. However, the headline result rests on a benchmark that is not released, for which no inter-annotator agreement or per-task sample sizes are reported, and whose design was itself motivated by failure-mode inspection on the same data; with single-run point estimates throughout, the significance of the stated accuracy gaps is not yet established.

major comments (3)
  1. [§III.A, Table III] The headline claim (abstract; Table III) that UAV-MAS-32B surpasses Gemini 3 Pro by 4.0% rests entirely on UAVQA-Bench, yet the benchmark is not released, is annotated by seven volunteers from the authors' own laboratory, and is not accompanied by any inter-annotator agreement statistic or per-task sample counts. Because the question templates (Appendix A) are fixed and known, the 1500 samples could over-represent task formats where the toolchain (detection, depth, zooming) is most beneficial; without a release, an external audit, or at minimum an IAA measure and per-task sample sizes, the 4.0% gap cannot be distinguished from benchmark-specific selection or template bias. I ask the authors to release the benchmark with annotation details, report IAA and per-task sizes, and discuss any potential template-driven biases in the composition statistics of Figure 3.
  2. [§IV, Tables IV–VI, Appendix B] The method was designed after evaluating failure modes on UAVQA-Bench itself (§IV states this explicitly), and the key hyperparameters — the S_init-to-τ mapping in §IV.C (τ=2,4,6), the depth D=5 and width W=min(3,|T_opt|) in Appendix B.A, and the sampling temperature of 0.7 — appear to be fixed values chosen for this same benchmark. All reported numbers are single-run point estimates with no standard deviation, error bars, or significance tests. This makes the claimed 7.4–8.7% gains over the base models and the 4.0% lead over Gemini 3 Pro vulnerable to overfitting to UAVQA-Bench. I request a hold-out tuning split or pre-registered hyperparameters, multiple seeds with variance reporting, and a sensitivity analysis over τ, D, and W to establish that the conclusions are not artifacts of these choices.
  3. [§V.D, Table VII] The CHOICE cross-dataset experiment is a real strength, but it provides only partial external validation: it tests the 8B model with generic zooming and description tools, not the full DSPE aerial toolkit, and it excludes Referring Expression Segmentation because of incompatible output formats, keeping only 440 compatible questions. Consequently, Table VII supports the generalizability of the CAIR/DAAS reasoning loop, but it does not directly validate the domain-specific toolset or the headline 32B-vs-Gemini-3-Pro comparison on UAVQA-Bench. The authors should state this limitation explicitly rather than implying full-system cross-dataset generalization, and ideally add a second external benchmark or a subset of CHOICE tasks with the complete DSPE toolkit to close the gap.
minor comments (5)
  1. [§IV.A] The de-hallucination filter in the Open-Vocabulary Detection tool uses an equidistant condition |c_{i+1}−2c_i+c_{i−1}|<δ, but the value of δ is never reported in the main text or the supplementary material; please specify it so the method is reproducible.
  2. [§IV.C] The S_init-to-τ mapping is described as 'fixed' but the boundaries [0,3], [4,8], [9,10] are discontinuous if S_init is real-valued; please clarify whether S_init is an integer score and, if so, state that explicitly.
  3. [§V.A, Table III] The 'Human Avg.' row reports the average of 10 participants but does not say how the participants were instructed, whether they used the same closed-ended protocol and image inputs as the models, or how much variance existed across participants; a brief protocol description and a standard deviation would be helpful.
  4. [Abstract and §V.D] The term 'training-free' is used for a system that relies on trained perception models such as Depth Anything 3 and open-vocabulary detectors; please clarify in the introduction or method that 'training-free' refers only to no fine-tuning of the MLLM/agent backbone, not to the absence of trained vision tools.
  5. [Throughout] There are inconsistent spacing artifacts in the UAV terminology (e.g., 'UA V-MAS' in the main text versus 'UAV-MAS' in tables and the supplementary); please harmonize the notation in the final camera-ready version.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity by construction; in-sample benchmark co-design is a validity risk, not a circular reduction.

full rationale

The paper's central claim is an empirical evaluation result, not a derivation constrained to equal its inputs. UAVQA-Bench is a newly constructed human-annotated benchmark, and UAV-MAS is a training-free agent system with fixed, explicitly stated components (DSPE, CAIR, DAAS). The DAAS threshold mapping is presented as a fixed three-level rule (S_init in [0,3],[4,8],[9,10] maps to tau = 2,4,6), not as a parameter fitted to the benchmark and then reported as a prediction. The statement that evaluating on UAVQA-Bench 'exposes three key failures' and that UAV-MAS was designed to address them does indicate in-sample adaptation, but no equation or construction makes the reported 77.0% accuracy equal to a fitted quantity; the accuracy is measured, not derived. Cross-dataset evaluation on CHOICE provides independent evidence that the method generalizes beyond UAVQA-Bench. No load-bearing self-citation, imported uniqueness theorem, or ansatz-by-citation is present. Accordingly, the appropriate finding is no significant circularity, and any concerns about benchmark representativeness or overfitting belong to correctness/validity risk rather than circularity.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities, forces, or dimensions; all new components are software agents and tools. The central claims rest on a set of hyperparameters (DAAS thresholds, depth, width, temperature) and domain assumptions about the reliability of off-the-shelf tools (Depth Anything 3, MLLM-based verification). The lack of released data means these assumptions cannot be independently checked.

free parameters (6)
  • DAAS pruning threshold mapping (tau) = tau=2 for S_init in [0,3], tau=4 for [4,8], tau=6 for [9,10]
    A fixed three-level mapping from initial score to pruning threshold is stated in Section IV-C. No sensitivity analysis is provided, and the values appear chosen to balance exploration on the authors' benchmark.
  • Maximum search depth D = 5
    Stated in Appendix B. This bound on DAAS tree depth is a hyperparameter that affects the accuracy-cost trade-off.
  • Search width W = min(3, |T_opt|)
    Stated in Appendix B. A small fixed width limits branching and is a design choice.
  • Sampling temperature for MLLM calls = 0.7
    Used for all MLLM calls, including CAIR sampling. Temperature controls the randomness of candidate generation in DAAS.
  • IoU threshold for visual grounding = 0.5
    A prediction is correct if IoU >= 0.5. This is a standard threshold but is a scoring choice that affects task difficulty.
  • De-hallucination filter delta = not specified
    The heuristic filter in DSPE removes detections when |c_{i+1} - 2c_i + c_{i-1}| < delta holds consecutively. The value of delta is not given, making the filter under-specified.
assumptions (4)
  • domain assumption Depth Anything 3 produces reliable relative depth maps for aerial imagery, allowing correct height and distance comparisons.
    Invoked in Section IV-A for the Distance Estimation tool. If the monocular depth prior fails on oblique or small-object drone views, the SRU tasks would be answered incorrectly.
  • domain assumption The MLLM-based existence verification and equidistant heuristic filter reliably separate real detections from hallucinated bounding boxes.
    Described in Section IV-A as the Open-Vocabulary Detection with De-hallucination tool. The entire DET tool depends on this two-stage verification working as intended.
  • domain assumption The Score Agent's self-reported confidence scores are meaningful for pruning and path selection in DAAS.
    DAAS uses S_init and per-step scores S' to decide branch pruning and optimal path selection. If these scores are noisy or miscalibrated, the search may prune good paths or select bad ones.
  • domain assumption The Instruct and Thinking variants of Qwen3-VL share the same architecture and can be mixed without breaking the system.
    Appendix B states that Instruct and Thinking models differ only in weights, not architecture, and mixes them (e.g., Agent LA uses Thinking). This is an unvalidated assumption about model compatibility.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Advancing MLLM-based UAV Image Understanding and Reasoning: A Benchmark and a Training-Free Multi-Agent System." pith.science (2026). https://pith.science/paper/ILHECK6T

@misc{pith2026260811738,
  author       = {Pith},
  title        = {Pith review of: Advancing MLLM-based UAV Image Understanding and Reasoning: A Benchmark and a Training-Free Multi-Agent System},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ILHECK6T}},
  note         = {Machine review of arXiv:2608.11738}
}
read the original abstract

Multimodal Large Language Model (MLLM)-based UAV aerial image understanding and reasoning is essential for aerial intelligence yet poses distinct challenges arising from extreme scale variation, arbitrary camera orientations, and high object density. Despite growing interest, existing evaluations remain fragmented across individual datasets and narrow tasks, leaving a critical gap in unified assessment of UAV understanding and reasoning capabilities. To fill this gap, we construct UAVQA-Bench, a benchmark of 1,500 human-annotated QA pairs drawn from 13 public UAV datasets, covering 6 capability dimensions and 16 tasks in both multiple-choice and visual grounding formats. Systematic evaluation of a broad range of open-source and closed-source MLLMs as well as agent-based systems on UAVQA-Bench identifies three key failure modes: domain-toolset mismatch, unchecked error propagation, and static reasoning. Motivated by these findings, we propose UAV-MAS, a training-free multi-agent system for MLLM-based UAV aerial image understanding and reasoning, comprising a Domain-Specific Perception Engine (DSPE) that routes queries to task-appropriate visual tools, a Context-Aware Iterative Refinement module (CAIR) that validates intermediate reasoning to curb error accumulation, and a Difficulty-Aware Adaptive Search mechanism (DAAS) that adjusts search depth to question difficulty. UAV-MAS with a 32B open-source MLLM achieves 77.0% overall accuracy on UAVQA-Bench, surpassing Gemini 3 Pro by 4.0\%, while the 8B variant improves 8.7\% over its base model.

Figures

Figures reproduced from arXiv: 2608.11738 by the authors.

Figure 1
Figure 1. Comparison of MLLM-based UAV aerial image understanding and [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of UAVQA-Bench. The benchmark assesses 6 key capabilities through 16 distinct tasks, collectively forming a comprehensive evaluation [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Dataset statistics of UAVQA-Bench. (a) Composition of the six capability dimensions. (b) Spatial distribution of target centers. (c) High-frequency [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Overview of UAV-MAS. (a) The overall pipeline of UAV-MAS. (b) The Domain-Specific Perception Engine (DSPE) tailored for UAV perception. (c) [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Visual comparison of UAV-MAS with other methods. All experiments are based on Qwen3-VL 8B except Gemini 3 Pro. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 1
Figure 1. Figure 1: Sample instances for tasks in Existence Detection, Category Recognition, and Quantity Awareness. [PITH_FULL_IMAGE:figures/full_fig_p016_1.png]
Figure 2
Figure 2. Figure 2: Sample instances for tasks in Fine-grained Attribute Perception and Spatial Relationship Understanding. [PITH_FULL_IMAGE:figures/full_fig_p017_2.png]
Figure 3
Figure 3. Figure 3: Sample instances for tasks in Visual Grounding. [PITH_FULL_IMAGE:figures/full_fig_p018_3.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

65 extracted references · 40 canonical work pages

  1. [1]

    Uav-detr: efficient end-to-end object detection for unmanned aerial vehicle imagery,

    H. Zhang, K. Liu, Z. Gan, and G.-N. Zhu, “Uav-detr: efficient end-to-end object detection for unmanned aerial vehicle imagery,”arXiv preprint arXiv:2501.01855, 2025

    arXiv 2025

  2. [2]

    Self-supervised monocular depth estimation from oblique uav videos,

    L. Madhuanand, F. Nex, and M. Y . Yang, “Self-supervised monocular depth estimation from oblique uav videos,”ISPRS journal of photogram- metry and remote sensing, vol. 176, pp. 1–14, 2021

    work page 2021

  3. [3]

    Mmbench: Is your multi-modal model an all-around player?

    Y . Liu, H. Duan, Y . Zhang, B. Li, S. Zhang, W. Zhao, Y . Yuan, J. Wang, C. He, Z. Liuet al., “Mmbench: Is your multi-modal model an all-around player?” inEuropean conference on computer vision. Springer, 2024, pp. 216–233

    2024

  4. [4]

    RSVQA: Visual question answering for remote sensing data,

    S. Lobry, D. Marcos, J. Murray, and D. Tuia, “RSVQA: Visual question answering for remote sensing data,”IEEE Transactions on Geoscience and Remote Sensing, vol. 58, no. 12, pp. 8555–8566, 2020

    work page 2020

  5. [5]

    Vrsbench: A versatile vision-language benchmark dataset for remote sensing image understanding,

    X. Li, J. Ding, and M. Elhoseiny, “Vrsbench: A versatile vision-language benchmark dataset for remote sensing image understanding,”Advances in Neural Information Processing Systems, vol. 37, pp. 3229–3242, 2024

    2024

  6. [6]

    Xlrs-bench: Could your multimodal llms understand extremely large ultra-high-resolution remote sensing im- agery?

    F. Wang, H. Wang, Z. Guo, D. Wang, Y . Wang, M. Chen, Q. Ma, L. Lan, W. Yang, J. Zhanget al., “Xlrs-bench: Could your multimodal llms understand extremely large ultra-high-resolution remote sensing im- agery?” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 14 325–14 336

    2025

  7. [7]

    Visdrone-det2019: The vision meets drone object detection in image challenge results,

    D. Du, P. Zhu, L. Wen, X. Bian, H. Lin, Q. Hu, T. Peng, J. Zheng, X. Wang, Y . Zhanget al., “Visdrone-det2019: The vision meets drone object detection in image challenge results,” inProceedings of the IEEE/CVF international conference on computer vision workshops, 2019, pp. 0–0

    2019

  8. [8]

    The unmanned aerial vehicle benchmark: Object detection and tracking,

    D. Du, Y . Qi, H. Yu, Y . Yang, K. Duan, G. Li, W. Zhang, Q. Huang, and Q. Tian, “The unmanned aerial vehicle benchmark: Object detection and tracking,” inProceedings of the European conference on computer vision (ECCV), 2018, pp. 370–386

    2018

Show all 65 references
  1. [9]

    Urbanvideo-bench: Benchmarking vision-language models on embodied intelligence with video data in urban spaces,

    B. Zhao, J. Fang, Z. Dai, Z. Wang, J. Zha, W. Zhang, C. Gao, Y . Wang, J. Cui, X. Chenet al., “Urbanvideo-bench: Benchmarking vision-language models on embodied intelligence with video data in urban spaces,”arXiv preprint arXiv:2503.06157, 2025

    2025

  2. [10]

    Yolo26: An analysis of nms-free end to end framework for real-time object detection,

    S. Chakrabarty, “Yolo26: An analysis of nms-free end to end framework for real-time object detection,”arXiv preprint arXiv:2601.12882, 2026. 12

    2026

  3. [11]

    Dfir-detr: Frequency domain enhancement and dynamic feature aggregation for cross-scene small object detection,

    B. Gao, J. Tong, X. Chen, H. Yu, and Z. Li, “Dfir-detr: Frequency domain enhancement and dynamic feature aggregation for cross-scene small object detection,”arXiv preprint arXiv:2512.07078, 2025

    2025 arXiv

  4. [12]

    Uav-based intelligent traffic surveillance system: Real-time vehicle detection, classification, tracking, and behav- ioral analysis,

    A. Khanpour, T. Wang, A. Vahidi-Shams, W. Ectors, F. Nakhaie, A. Taheri, and C. Claudel, “Uav-based intelligent traffic surveillance system: Real-time vehicle detection, classification, tracking, and behav- ioral analysis,”arXiv preprint arXiv:2509.04624, 2025

    2025 arXiv

  5. [13]

    Global–local fusion with semantic information-guidance for accurate small object detection in UA V aerial images,

    Y . Chen, Z. Ye, H. Sun, T. Gong, S. Xiong, and X. Lu, “Global–local fusion with semantic information-guidance for accurate small object detection in UA V aerial images,”IEEE Transactions on Geoscience and Remote Sensing, vol. 63, pp. 1–15, 2025, art. no. 4701115

    2025

  6. [14]

    Detection-driven exposure-correction network for nighttime drone-view object detection,

    Y . Xi, W. Jia, Q. Miao, J. Feng, J. Ren, and H. Luo, “Detection-driven exposure-correction network for nighttime drone-view object detection,” IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1– 14, 2024, art. no. 5605014

    2024

  7. [15]

    Wheatai v1. 0: An ai-powered high throughput wheat phenotyping platform,

    M. Maimaitijiang, H. Ghimire, S. Thapa, M. M. Billah, S. Sehgal, M. Singh, S. Kaushal, K. Poudel, S. Subedi, U. U. R. Janjuaet al., “Wheatai v1. 0: An ai-powered high throughput wheat phenotyping platform,”arXiv preprint arXiv:2601.08863, 2026

    2026

  8. [16]

    Maizestandcounting (masc): Auto- mated and accurate maize stand counting from uav imagery using image processing and deep learning,

    D. E. Kharismawati and T. Kazic, “Maizestandcounting (masc): Auto- mated and accurate maize stand counting from uav imagery using image processing and deep learning,”arXiv preprint arXiv:2510.07580, 2025

    2025

  9. [17]

    Panoptic segmentation of environmental uav images: Litter beach,

    O. Youme, J. M. Dembele, E. C. Ezin, and C. Cambier, “Panoptic segmentation of environmental uav images: Litter beach,”arXiv preprint arXiv:2508.15985, 2025

    2025 arXiv

  10. [18]

    Codrone: Autonomous drone navigation assisted by edge and cloud foundation models,

    P. Chen, T. Ouyang, K. Luo, W. Hong, and X. Chen, “Codrone: Autonomous drone navigation assisted by edge and cloud foundation models,”IEEE Internet of Things Journal, 2025

    2025

  11. [19]

    Generalization evaluation of deep stereo matching methods for uav-based forestry applications,

    Y . Lin, B. Xue, M. Zhang, S. Schofield, and R. Green, “Generalization evaluation of deep stereo matching methods for uav-based forestry applications,”arXiv preprint arXiv:2512.03427, 2025

    2025

  12. [20]

    Precise depth estimation by calculating affine transformation parameters,

    X. Shi, T. Tang, J. Chen, S. Lv, and Y . Liu, “Precise depth estimation by calculating affine transformation parameters,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–15, 2024

    2024

  13. [21]

    FLDet: Faster and lighter aerial object detector,

    S. Wang, K. Liu, J. Huang, and X. Li, “FLDet: Faster and lighter aerial object detector,”IEEE Transactions on Circuits and Systems for Video Technology, vol. 35, no. 5, pp. 4450–4463, 2025

    2025

  14. [22]

    TGCADNet: Text-guided context-aware detection via CLIP for small objects in UA V scenes,

    F. Sun, D. Cheng, P. Zheng, T. Song, L. Chen, and Q. Kou, “TGCADNet: Text-guided context-aware detection via CLIP for small objects in UA V scenes,”IEEE Transactions on Circuits and Systems for Video Technology, vol. 36, no. 6, pp. 7846–7859, 2026

    2026

  15. [23]

    Detect anything via next point prediction,

    Q. Jiang, J. Huo, X. Chen, Y . Xiong, Z. Zeng, Y . Chen, T. Ren, J. Yu, and L. Zhang, “Detect anything via next point prediction,”arXiv preprint arXiv:2510.12798, 2025

    2025

  16. [24]

    Visdrone-det2021: The vision meets drone object detection challenge results,

    Y . Cao, Z. He, L. Wang, W. Wang, Y . Yuan, D. Zhang, J. Zhang, P. Zhu, L. Van Gool, J. Hanet al., “Visdrone-det2021: The vision meets drone object detection challenge results,” inProceedings of the IEEE/CVF International conference on computer vision, 2021, pp. 2847–2854

    2021

  17. [25]

    Dota: A large-scale dataset for object detection in aerial images,

    G.-S. Xia, X. Bai, J. Ding, Z. Zhu, S. Belongie, J. Luo, M. Datcu, M. Pelillo, and L. Zhang, “Dota: A large-scale dataset for object detection in aerial images,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 3974–3983

    2018

  18. [26]

    Uavscenes: A multi-modal dataset for uavs,

    S. Wang, S. Li, Y . Zhang, S. Yu, S. Yuan, R. She, Q. Guo, J. Zheng, O. K. Howe, L. Chandraet al., “Uavscenes: A multi-modal dataset for uavs,”arXiv preprint arXiv:2507.22412, 2025

    2025 arXiv

  19. [27]

    Dyfo: A training-free dynamic focus visual search for enhancing lmms in fine-grained visual understanding,

    G. Li, J. Xu, Y . Zhao, and Y . Peng, “Dyfo: A training-free dynamic focus visual search for enhancing lmms in fine-grained visual understanding,” inProceedings of the Computer Vision and Pattern Recognition Confer- ence, 2025, pp. 9098–9108

    2025

  20. [28]

    Pyvision: Agentic vision with dynamic tooling,

    S. Zhao, H. Zhang, S. Lin, M. Li, Q. Wu, K. Zhang, and C. Wei, “Pyvision: Agentic vision with dynamic tooling,”arXiv preprint arXiv:2507.07998, 2025

    2025 arXiv

  21. [29]

    Multi-agent constraint factorization reveals latent invariant solution structure,

    C. Scofield, “Multi-agent constraint factorization reveals latent invariant solution structure,”arXiv preprint arXiv:2601.15077, 2026

    2026

  22. [30]

    Mas-orchestra: Understanding and improv- ing multi-agent reasoning through holistic orchestration and controlled benchmarks,

    Z. Ke, Y . Ming, A. Xu, R. Chin, X.-P. Nguyen, P. Jwalapuram, S. Yavuz, C. Xiong, and S. Joty, “Mas-orchestra: Understanding and improv- ing multi-agent reasoning through holistic orchestration and controlled benchmarks,”arXiv preprint arXiv:2601.14652, 2026

    2026 arXiv

  23. [31]

    Agent identity uri scheme: Topology-independent naming and capability-based discovery for multi-agent systems,

    R. R. Rodriguez Jr, “Agent identity uri scheme: Topology-independent naming and capability-based discovery for multi-agent systems,”arXiv preprint arXiv:2601.14567, 2026

    2026 arXiv

  24. [32]

    Internvl3. 5: Advancing open-source multi- modal models in versatility, reasoning, and efficiency,

    W. Wang, Z. Gao, L. Gu, H. Pu, L. Cui, X. Wei, Z. Liu, L. Jing, S. Ye, J. Shaoet al., “Internvl3. 5: Advancing open-source multi- modal models in versatility, reasoning, and efficiency,”arXiv preprint arXiv:2508.18265, 2025

    2025 arXiv

  25. [33]

    Pa- per2rebuttal: A multi-agent framework for transparent author response assistance,

    Q. Ma, C. Guo, Z. Tian, S. Wang, J. Xiao, Y . Yue, and Z. Zhang, “Pa- per2rebuttal: A multi-agent framework for transparent author response assistance,”arXiv preprint arXiv:2601.14171, 2026

    2026 arXiv

  26. [34]

    The orchestration of multi- agent systems: Architectures, protocols, and enterprise adoption,

    A. Adimulam, R. Gupta, and S. Kumar, “The orchestration of multi- agent systems: Architectures, protocols, and enterprise adoption,”arXiv preprint arXiv:2601.13671, 2026

    2026

  27. [35]

    Motion-to-response content generation via multi-agent ai sys- tem with real-time safety verification,

    H. Lee, “Motion-to-response content generation via multi-agent ai sys- tem with real-time safety verification,”arXiv preprint arXiv:2601.13589, 2026

    2026

  28. [36]

    Agentgc: Evolutionary learning-based lossless compres- sion for genomics data with llm-driven multiple agent,

    S. Hui, D. Yanfeng, H. Ma, C. Xu, K. Jin, L. Zu, C. Zhong, G. Wang, W. Caiet al., “Agentgc: Evolutionary learning-based lossless compres- sion for genomics data with llm-driven multiple agent,”arXiv preprint arXiv:2601.13559, 2026

    2026

  29. [37]

    React: Synergizing reasoning and acting in language models,

    S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. R. Narasimhan, and Y . Cao, “React: Synergizing reasoning and acting in language models,” inThe eleventh international conference on learning representations, 2022

    2022

  30. [38]

    Au-air: A multi-modal unmanned aerial vehicle dataset for low altitude traffic surveillance,

    I. Bozcan and E. Kayacan, “Au-air: A multi-modal unmanned aerial vehicle dataset for low altitude traffic surveillance,” in2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 8504–8510

    2020

  31. [39]

    Webuav-3m: A benchmark for unveiling the power of million- scale deep uav tracking,

    C. Zhang, G. Huang, L. Liu, S. Huang, Y . Yang, X. Wan, S. Ge, and D. Tao, “Webuav-3m: A benchmark for unveiling the power of million- scale deep uav tracking,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 7, pp. 9186–9205, 2022

    2022

  32. [40]

    Visdrone-det2019: The vision meets drone object detection in image challenge results,

    D. Du, P. Zhu, L. Wen, X. Bian, H. Lin, Q. Hu, T. Peng, J. Zheng, X. Wang, Y . Zhanget al., “Visdrone-det2019: The vision meets drone object detection in image challenge results,” inProceedings of the IEEE/CVF international conference on computer vision workshops, 2019

    2019

  33. [41]

    Semantic drone dataset,

    U. Graz, “Semantic drone dataset,” 2019. [Online]. Available: http://dronedataset.icg.tugraz.at/

    2019

  34. [42]

    Drone-based rgb-infrared cross- modality vehicle detection via uncertainty-aware learning,

    Y . Sun, B. Cao, P. Zhu, and Q. Hu, “Drone-based rgb-infrared cross- modality vehicle detection via uncertainty-aware learning,”IEEE Trans- actions on Circuits and Systems for Video Technology, vol. 32, no. 10, pp. 6700–6713, 2022

    2022

  35. [43]

    Vdd: Varied drone dataset for semantic segmentation,

    W. Cai, K. Jin, J. Hou, C. Guo, L. Wu, and W. Yang, “Vdd: Varied drone dataset for semantic segmentation,”Journal of Visual Communication and Image Representation, vol. 109, p. 104429, 2025

    2025

  36. [44]

    Large-scale structure from motion with semantic constraints of aerial images,

    Y . Chen, Y . Wang, P. Lu, Y . Chen, and G. Wang, “Large-scale structure from motion with semantic constraints of aerial images,” inChinese Conference on Pattern Recognition and Computer Vision (PRCV). Springer, 2018, pp. 347–359

    2018

  37. [45]

    Uavid: A semantic segmentation dataset for uav imagery,

    Y . Lyu, G. V osselman, G.-S. Xia, A. Yilmaz, and M. Y . Yang, “Uavid: A semantic segmentation dataset for uav imagery,”ISPRS journal of photogrammetry and remote sensing, vol. 165, pp. 108–119, 2020

    2020

  38. [46]

    Wilduav: Monocular uav dataset for depth estimation tasks,

    H. Florea, V .-C. Miclea, and S. Nedevschi, “Wilduav: Monocular uav dataset for depth estimation tasks,” in2021 IEEE 17th International Conference on Intelligent Computer Communication and Processing (ICCP). IEEE, 2021, pp. 291–298

    2021

  39. [47]

    Hazydet: Open-source benchmark for drone-view object detec- tion with depth-cues in hazy scenes,

    C. Feng, Z. Chen, X. Li, C. Wang, J. Yang, M.-M. Cheng, Y . Dai, and Q. Fu, “Hazydet: Open-source benchmark for drone-view object detec- tion with depth-cues in hazy scenes,”arXiv preprint arXiv:2409.19833, 2024

    2024 arXiv

  40. [48]

    Graph regularized flow attention network for video animal counting from drones,

    P. Zhu, T. Peng, D. Du, H. Yu, L. Zhang, and Q. Hu, “Graph regularized flow attention network for video animal counting from drones,”IEEE Transactions on Image Processing, vol. 30, pp. 5339–5351, 2021

    2021

  41. [49]

    A benchmark and simulator for uav tracking,

    M. Mueller, N. Smith, and B. Ghanem, “A benchmark and simulator for uav tracking,” inEuropean conference on computer vision, vol. 7, 2016

    2016

  42. [50]

    Depth anything 3: Recovering the visual space from any views,

    H. Lin, S. Chen, J. H. Liew, D. Y . Chen, Z. Li, G. Shi, J. Feng, and B. Kang, “Depth anything 3: Recovering the visual space from any views,”arXiv preprint arXiv:2511.10647, 2025

    2025 arXiv

  43. [51]

    Qwen3-vl technical report,

    S. Bai, Y . Cai, R. Chen, K. Chen, X. Chen, Z. Cheng, L. Deng, W. Ding, C. Gao, C. Ge, W. Ge, Z. Guo, Q. Huang, J. Huang, F. Huang, B. Hui, S. Jiang, Z. Li, M. Li, M. Li, K. Li, Z. Lin, J. Lin, X. Liu, J. Liu, C. Liu, Y . Liu, D. Liu, S. Liu, D. Lu, R. Luo, C. Lv, R. Men, L. M...

    2025 arXiv

  44. [52]

    Glm-4.5v and glm-4.1v-thinking: Towards versatile multimodal reasoning with scalable reinforcement learning,

    V . Team, W. Hong, W. Yu, X. Gu, G. Wang, G. Gan, H. Tang, J. Cheng, J. Qi, J. Ji, L. Pan, S. Duan, W. Wang, Y . Wang, Y . Cheng, Z. He, Z. Su, Z. Yang, Z. Pan, A. Zeng, B. Wang, B. Chen, B. Shi, C. Pang, C. Zhang, D. Yin, F. Yang, G. Chen, J. Xu, J. Zhu, J. Chen, J. Chen, J. ...

    2025 arXiv

  45. [53]

    Qwen3.5: Accelerating productivity with native multimodal agents,

    Q. Team, “Qwen3.5: Accelerating productivity with native multimodal agents,” February 2026. [Online]. Available: https://qwen.ai/blog?id= qwen3.5

    2026

  46. [54]

    Introducing gpt-5.2,

    OpenAI, “Introducing gpt-5.2,” https://openai.com/index/ introducing-gpt-5-2/, 2025, accessed: 2026-01-28

    2025

  47. [55]

    Gemini 3 pro and gemini 3 flash models,

    Google DeepMind, “Gemini 3 pro and gemini 3 flash models,” https: //ai.google.dev/, 2025, accessed: 2026-01-15

    2025

  48. [56]

    Qwen-agent: An agent framework based on qwen,

    QwenLM, “Qwen-agent: An agent framework based on qwen,” https: //github.com/QwenLM/Qwen-Agent, 2024

    2024

  49. [57]

    Choice: Benchmarking the remote sensing capabilities of large vision-language models,

    X. An, J. Sun, Z. Gui, and W. He, “Choice: Benchmarking the remote sensing capabilities of large vision-language models,” in Advances in Neural Information Processing Systems, vol. 38, 2025. [Online]. Available: https://proceedings.neurips.cc/paper files/paper/ 2025/hash/befe2...

    2025

  50. [59]

    Existence Detection (ED): •Scene Presence ◦Is any{object}present here? ◦Is any{object}visible in the image? ◦Is there any{object}in the scene? ◦Does this image contain any{object}? ◦Can you see any{object}in this picture? •Conditional Presence ◦Is any{object}present{condition}...

  51. [60]

    ◦Classify the object in region{<x 1><y1><x2><y2>}

    Category Recognition (CR): •Regional Classification ◦Identify the object found in{<x 1><y1><x2><y2>}. ◦Classify the object in region{<x 1><y1><x2><y2>}. ◦Which class does the object in{<x 1><y1><x2><y2>}belong to? ◦What is the category of the object located at{<x 1><y1><x2><y2...

  52. [61]

    ◦Count all the{object}in this picture

    Quantity Awareness (QA): •Scene Counting ◦Provide the quantity of{object}. ◦Count all the{object}in this picture. ◦How many{object}are in this image? ◦How many instances of{object}can you see? ◦Can you count the number of{object}present? •Regional Counting ◦Provide the quantit...

  53. [62]

    ◦Which color does the{object}in{<x 1><y1><x2><y2>}have? ◦Identify the driving direction (relative to the camera) of the{object}in{<x 1><y1><x2><y2>}

    Fine-grained Attribute Perception (FAP): •Regional Attribute Recognition ◦What color is the{object}in{<x 1><y1><x2><y2>}? ◦Identify the color of the{object}in{<x 1><y1><x2><y2>}. ◦Which color does the{object}in{<x 1><y1><x2><y2>}have? ◦Identify the driving direction (relative ...

  54. [63]

    Spatial Relationship Understanding (SRU): •Spatial Relation ◦Relative to the object in{<x 1><y1><x2><y2>}A, where is the object in{<x 1><y1><x2><y2>}B? ◦As seen by the object in{<x 1><y1><x2><y2>}A, in what direction is the object in{<x 1><y1><x2><y2>}B? ◦From the viewpoint of...

  55. [64]

    Yes", "No

    Visual Grounding (VG): •Simple Object Grounding ◦Locate the{object}. ◦Point out the{object}. ◦Could you point out the{object}? ◦Draw the position of a/an{object}. 3 ◦Can you show me where the{object}is? •Complex Semantic Grounding ◦Please give me the location of{semantics}. ◦P...

  56. [65]

    Q”, “O”, and “A

    Template Visualizations:In this section, we present additional examples from UA VQA-Bench. Figure 1 provides sample instances and visual prompts for the tasks of Existence Detection, Category Recognition, and Quantity Awareness. Figure 2 illustrates examples for Fine-grained A...

  57. [2013]

    He is currently a Professor with the College of Future In- formation Technology, Fudan University, Shanghai, China

    He was a Research Scientist with the Institute for Infocomm Research, A*STAR, Singapore, from 2013 to 2017, and a Senior Scientist with the Huawei Singapore Research Center from 2017 to 2018. He is currently a Professor with the College of Future In- formation Technology, Fuda...

    2013

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.