Writing study: DroneDINO: Towards Heterogeneous Routed Mixture of Experts for Drone-based Unified Object Detection

ICML overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← ICML 2026 oral writing corpus

Dongdong Li, Rui Chen, Yan Fan, Yan Liu, Yangliu Kuai, Pengfei Zhu · Official ICML 2026 oral record

Observed writing choices, not acceptance rules. This study examines the inspected manuscript version. It does not establish why the paper was accepted or that its presentation choices should be copied unchanged.

Figure, table and section ledger

Editorial contribution type: Architecture & representation (primary).

Layout extraction, not a required paper template. Counts are distinct caption labels; a multi-panel figure counts once. Section spans are estimated from heading positions, including floats and whitespace. Method and limitations.

Canonical PDF

10 physical pages

Before references / appendix

≈ 8.1 page-equivalents

Main figures / tables

4 / 7

Attached appendix figures / tables

0 / 0

Attached appendix span

No boundary detected; not assumed zero

Before references / appendix: ≈ 81%References / intervening material: ≈ 19%Estimated layout envelopes, not text/artwork area. Unresolved appendix boundaries remain within the undivided post-body material.

Extraction notes: No attached appendix boundary detected. Zero means no recognized caption, not independently verified absence.

Section-by-section lengths and source pages

The title/abstract/front matter precede the first numbered heading. A span ends at the next detected top-level heading, not at the last sentence; end-page whitespace is included. References and statements are listed separately. These are not measurements of text-only or artwork area.

Top-level heading	Starts on PDF page	Estimated span	Share of entire PDF
1. Introduction	1	≈ 1.4 pp.	14%
2. Related Work	2	≈ 0.6 pp.	6%
3. Method	3	≈ 2.42 pp.	24.2%
4. Experiments	5	≈ 2.93 pp.	29.3%
5. Conclusion and Limitation	8	≈ 0.31 pp.	3.1%
Impact Statement	9	≈ 0.05 pp.	0.5%
References	9	≈ 1.9 pp.	19%

All detected figure and table locations

Main text

• Figure 1 · p. 1
• Table 1 · p. 3
• Figure 2 · p. 4
• Table 2 · p. 6
• Figure 3 · p. 6
• Table 3 · p. 7
• Table 4 · p. 7
• Table 5 · p. 7
• Table 6 · p. 7
• Figure 4 · p. 8
• Table 7 · p. 8

Attached appendix

No recognized caption labels.

Argument and organization

1. 1 Introduction; 2 Related Work

   Defines the cross-dataset/modalities problem and compares unified detection with specialist models.

2. 3 Method

   Enumerates five subtasks, a shared interface, heterogeneous experts and training-only task recognition.

3. 4 Experiments; 5 Conclusion and Limitation

   Separates unified and specialist comparisons, component/routing ablations and restricted output/task scope.

Problem, claim and contribution

The opening links heterogeneous drone datasets to imbalanced expert activation and presents shared, task-specific and dynamic experts as distinct responses.

Abstract, p.1; task definition, p.3 · Main paper

Argument and section order

A task table defines modality, targets and source dataset before the unified token equation, making the scope of unification concrete.

Section 3.1 and Table 1, p.3 · Main paper

Figures, captions and visual explanation

The architecture figure separates the complete detector from an enlarged expert-routing module; later heatmaps align original images and with/without-module outputs in rows.

Figure 2, p.4; Figure 4, p.8 · Main paper

Tables, comparisons and evidence

Unified baselines are compared separately from specialist detectors, and the unified table retains parameter/FLOP costs and unavailable fusion cells rather than reporting only accuracy.

Table 2, p.6; Tables 3-5, p.7 · Main paper

Appendix and reproducibility

With no appendix in the retrieved ten-page PDF, implementation settings and expert-count/routing ablations are located in the main text.

Section 4.2, p.6; Table 7, p.8; PDF ends p.10 · Main paper

Clarity and writing conventions

The conclusion explicitly limits the model to horizontal boxes and multiple dataset-specific heads, while the result text acknowledges that UAV-DETR is stronger on VisDrone.

Section 4.4, p.7; Section 5, p.8 · Main paper

What to reuse—and what not to infer

Transferable lesson: Define what a unified model actually shares, and separate broader input coverage from accuracy and computational-cost claims.

Boundary: The architecture has higher listed parameters/FLOPs than unified baselines, does not win every specialist comparison, and selected activation maps are illustrative rather than causal verification.

Apply this lesson using the argument and evidence worksheet.

Inspection record and sources

Checked

2026-10-07

Reading status

Writing-focused manuscript scan + selected visuals inspected

Main text

Writing-oriented scan of the main argument through physical p.8, with focused physical-page text inspection on 1, 3, 5, 6, 7, 8. Framing, section progression, mechanism/proof explanation, displays, comparison conditions and conclusions were inspected. This is not exhaustive technical reading, proof verification or reproduction.

Appendix / supplement

The retrieved proceedings PDF has 10 physical pages, with main text through p.8 and references on pp.9-10. No embedded appendix was present and no separately linked supplement was inspected.

Visually inspected PDF pages

4, 8

This is a writing-focused scan of the manuscript and the stated supplementary scope: framing, organization, visual explanation, evidence presentation and wording. It is not a line-by-line proof audit, reproduction of results or exhaustive scientific review.

• Published proceedings PDF, PMLR volume 306, retrieved 2026-10-07 · 10 pages