Writing study: LayerTracer: Cognitive-Aligned Layered SVG Synthesis via Diffusion Transformer

ICCV overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← ICCV 2025 oral writing corpus

Yiren Song, Danze Chen, Mike Zheng Shou · Official ICCV 2025 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: Domain-grounded methods & physical systems (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

11 physical pages

Before references / appendix

≈ 8.05 page-equivalents

Main figures / tables

9 / 3

Attached appendix figures / tables

0 / 0

Attached appendix span

No boundary detected; not assumed zero

Before references / appendix: ≈ 73.18%References / intervening material: ≈ 26.82%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	≈ 0.89 pp.	8.09%
2. Related Works	2	≈ 0.96 pp.	8.73%
3. Method	3	≈ 2.12 pp.	19.27%
4. Experiment	5	≈ 2.95 pp.	26.82%
5. Limitations and Future Work	8	≈ 0.06 pp.	0.55%
6. Conclusion	8	≈ 0.15 pp.	1.36%
Acknowledgement	8	≈ 0.13 pp.	1.18%
References	9	≈ 2.95 pp.	26.82%

All detected figure and table locations

Main text

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

Attached appendix

No recognized caption labels.

Argument and organization

1. 1 Introduction; 2 Related Works

   Motivates editable semantic layers rather than only a convincing rasterized SVG.

2. 3 Method

   Constructs layer sequences from SVG hierarchy, trains text/image-conditioned diffusion and then vectorizes per-layer raster differences.

3. 4 Experiment; 5 Limitations and Future Work; 6 Conclusion

   Separates text generation, vectorization, serpentine-layout ablation and human layer-rationality preference.

Problem, claim and contribution

The paper treats path redundancy and incoherent semantic grouping as editability problems that image resemblance alone misses.

Sections 1–2, physical pp.1–3 · Main paper

Argument and section order

Data sequencing, raster generation and final vectorization have separate subsections, exposing the learned and conventional components of the pipeline.

Sections 3.3–3.6, pp.4–5 · Main paper

Figures, captions and visual explanation

The overview maps serpentine raster grids to two LoRA training stages and layer-wise vectorization; the ablation shows repeated/inconsistent frames without the layout strategy.

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

Tables, comparisons and evidence

Evaluation distinguishes 50 text prompts from 50 vectorization inputs, while the 46-participant study asks separately about preference, prompt match and layer rationality.

Sections 4.1 and 4.5, pp.6,8 · Main paper

Appendix and reproducibility

Additional experimental results are assigned to supplementary material; this scan inspected main rasterized layer examples, not editable output SVGs or supplemental results.

Section 4.2, p.7 · Main paper

Clarity and writing conventions

The limitations identify dependence on Vtracer/manual hyperparameter adjustment and failure on complex out-of-distribution inputs.

Section 5, p.8 · Main paper

What to reuse—and what not to infer

Transferable lesson: Expose the learned raster-to-vector handoff and evaluate semantic layer organization separately from the final rendered image.

Boundary: Sequences reconstructed from SVG grouping are not recorded human design actions or direct evidence of cognitive alignment. Raster examples and small benchmark/preference sets do not establish professional editability; vectorization still inherits conventional tracing and tuning limitations.

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 across the main manuscript, physical PDF pp.1–8, covering framing, section progression, method explanations, captions, comparison protocols and concluding scope; reference boundary checked. Selected full-page rasters were inspected. Not exhaustive line-by-line reading, proof verification or experimental reproduction.

Appendix / supplement

Main-paper appendix/supplement references were checked. A separate supplementary source is linked in the inventory but was not retrieved or inspected; its contents are not included in the reading or visual-page counts.

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.

• Official CVF open-access ICCV 2025 accepted-paper PDF, retrieved 7 October 2026; CVF states identical to accepted version except watermark; not assumed identical to initial submission · 11 pages