Writing study: Monocular Normal Estimation via Shading Sequence Estimation

ICLR overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← ICLR 2026 oral writing corpus

Zongrui Li, Xinhua Ma, Minghui HU, Yunqing Zhao, Yingchen Yu, Qian Zheng, Chang Liu, Jiang Xudong, Song Bai · Official ICLR 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 design (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

13 physical pages

Before references / appendix

≈ 10.11 page-equivalents

Main figures / tables

6 / 6

Attached appendix figures / tables

0 / 0

Attached appendix span

Separate file recorded below; not in this PDF total

Before references / appendix: ≈ 77.77%References / intervening material: ≈ 22.23%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	2	≈ 1.06 pp.	8.15%
2 RELATED WORKS	3	≈ 0.76 pp.	5.85%
3 METHODS	4	≈ 2.39 pp.	18.38%
4 EXPERIMENTS	6	≈ 3.81 pp.	29.31%
5 DISCUSSION	10	≈ 0.57 pp.	4.38%
REFERENCES	11	≈ 2.89 pp.	22.23%

All detected figure and table locations

Main text

• Figure 1 · p. 1
• Figure 2 · p. 2
• Figure 3 · p. 2
• Figure 5 · p. 5
• Figure 4 · p. 5
• Table 1 · p. 7
• Table 2 · p. 7
• Figure 6 · p. 8
• Table 3 · p. 8
• Table 4 · p. 9
• Table 5 · p. 9
• Table 6 · p. 10

Attached appendix

No recognized caption labels.

Separate source files—excluded from canonical-PDF totals

A separate author version may repeat main text and references. Whole-file lengths and counts are not extra appendix material and are never added to the canonical PDF totals. Read the study's source scope for which sections received substantive reading versus layout inspection.

Additional source 1

Separate PDF (or ZIP containing it): whole-file layout of 28 physical pages; 25 detected figures and 11 tables. These may include overlapping main-text material.

Detected heading	PDF page	Estimated span

No top-level headings detected; no section lengths inferred.

• Table 7 · p. 2
• Table 8 · p. 3
• Figure 7 · p. 3
• Figure 8 · p. 4
• Figure 9 · p. 5
• Figure 10 · p. 6
• Figure 11 · p. 7
• Figure 12 · p. 8
• Figure 13 · p. 9
• Figure 14 · p. 10
• Figure 15 · p. 11
• Figure 16 · p. 11
• Figure 17 · p. 12
• Figure 18 · p. 12
• Figure 19 · p. 13
• Figure 20 · p. 14
• Figure 21 · p. 15
• Figure 22 · p. 15
• Figure 23 · p. 16
• Figure 24 · p. 17
• Figure 25 · p. 18
• Figure 26 · p. 19
• Figure 27 · p. 20
• Table 9 · p. 21
• Table 10 · p. 22
• Table 11 · p. 23
• Figure 28 · p. 24
• Figure 29 · p. 24
• Figure 30 · p. 25
• Figure 31 · p. 26
• Table 12 · p. 26
• Table 13 · p. 27
• Table 14 · p. 27
• Table 15 · p. 27
• Table 16 · p. 28
• Table 17 · p. 28

Argument and organization

1. Failure case, alternative target, analytic recovery

   Explains why visually plausible normals can still reconstruct the wrong geometry.

2. Real and synthetic benchmarks, ablations, limitations

   Tests target, data, conditioning and backbone choices separately.

Problem, claim and contribution

Figures 1–3 distinguish normal-map appearance from recovered geometry, so the proposed intermediate shading target addresses a visible failure rather than a vaguely defined accuracy gap.

Introduction, pages 1–3. · Main paper · Additional source 2

Argument and section order

Named ablation paragraphs connect each design choice to a comparison, including retraining a baseline on MultiShade to separate dataset and method effects.

Section 4.4, pages 9–10; Table 6. · Main paper · Additional source 2

Figures, captions and visual explanation

The pipeline uses actual object, grayscale, shading and normal images; lock/fire icons clearly identify frozen encoders and the trainable diffusion component.

Figure 5, page 5. · Main paper · Additional source 2

Tables, comparisons and evidence

The per-object tables show both overall improvements and exceptions. The prose explicitly acknowledges weaker Goblet and House cases instead of claiming a universal win.

Tables 1–2 and Section 4.2, page 7. · Main paper · Additional source 2

Appendix and reproducibility

The supplement supplies selected test-image indices, runtime, training settings and representative-image selection rules, followed by controlled texture/material/light comparisons.

Supplement C, F.1–F.2 and G, printed pages 15, 19–23 and 37–38. · Main paper · Additional source 2

Clarity and writing conventions

The limitations distinguish computational overhead, dark regions, transparency and object-versus-scene scope; these are concrete boundaries useful to a potential adopter.

Section 5, page 10. · Main paper · Additional source 2

What to reuse—and what not to infer

Transferable lesson: Show the failure in the representation the application actually uses, then connect a visually explicit intermediate target to an analytic final estimator.

Boundary: Do not equate generated shading with measured photometric-stereo input. Lossless recovery assumes sufficient positive, independent lighting equations; the six-light guarantee needs care at boundary normals, and evaluated scene generality is limited.

Apply this lesson using the argument and evidence worksheet.

Inspection record and sources

Checked

2026-09-19

Reading status

Full text + selected visuals inspected

Main text

Main PDF pages 1–10, Sections 1–5; references on pages 11–13.

Appendix / supplement

Separate supplemental PDF pages 1–28 (printed pages 14–41), Appendices A–H, including implementation, all result captions and additional analyses.

Visually inspected PDF pages

5, 7

• Official ICLR 2026 proceedings conference PDF; PDF creation metadata: undefined · 13 pages
• Official ICLR 2026 separate supplemental PDF, downloaded 2026-09-19; printed pages 14–41. · 28 pages