Writing study: SparseFlex: High-Resolution and Arbitrary-Topology 3D Shape Modeling

ICCV overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← ICCV 2025 oral writing corpus

Xianglong He, Zi-Xin Zou, Chia-Hao Chen, Yuan-Chen Guo, Ding Liang, Chun Yuan, Wanli Ouyang, Yan-Pei Cao, Yangguang Li · 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: 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

12 physical pages

Before references / appendix

≈ 8.1 page-equivalents

Main figures / tables

6 / 5

Attached appendix figures / tables

0 / 0

Attached appendix span

No boundary detected; not assumed zero

Before references / appendix: ≈ 67.5%References / intervening material: ≈ 32.5%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	≈ 0.92 pp.	7.67%
2. Related Work	2	≈ 1.11 pp.	9.25%
3. Method	3	≈ 2.63 pp.	21.92%
4. Experiments	6	≈ 2.23 pp.	18.58%
5. Conclusion	8	≈ 0.27 pp.	2.25%
6. Acknowledgement	9	≈ 0.06 pp.	0.5%
References	9	≈ 3.9 pp.	32.5%

All detected figure and table locations

Main text

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

Attached appendix

No recognized caption labels.

Argument and organization

1. 1 Introduction; 2 Related Work

   Identifies watertight preprocessing and dense high-resolution extraction as separate representation bottlenecks.

2. 3 Method

   Builds a sparse differentiable surface representation, VAE, sectional rendering supervision and conditional flow generation.

3. 4 Experiments; 5 Conclusion

   Separates reconstruction from generation, full datasets from watertight subsets and representation sparsity from training-frustum savings.

Problem, claim and contribution

Open surfaces and interior structures are treated as specific topology requirements rather than subsumed under a generic higher-fidelity claim.

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

Argument and section order

Representation, autoencoder, training strategy and generation model each receive a separate subsection, avoiding conflation of reconstruction capacity and generative quality.

Sections 3.1–3.4, pp.4–6 · Main paper

Figures, captions and visual explanation

The overview connects voxelized point features through compression and self-pruning to surface rendering; the sectional diagram explains which voxels are activated.

Figure 2, p.3; Figure 3, p.4 · Main paper

Tables, comparisons and evidence

Reconstruction tables distinguish all assets from watertight subsets, while the resource table explicitly varies resolution and visible-voxel ratio and marks out-of-memory entries.

Table 1, p.6; Table 3, p.8 · Main paper

Appendix and reproducibility

Precise rendering-loss definitions and weighting coefficients are assigned to supplementary material, which was not inspected.

Section 3.3, p.5 · Main paper

Clarity and writing conventions

Limitations distinguish low-resolution boundary artifacts, high-resolution computational demand and limited control of generated interiors.

Conclusion, p.8 · Main paper

What to reuse—and what not to infer

Transferable lesson: Keep reconstruction and generation evaluations separate, and report resource savings against resolution and visibility rather than one headline number.

Boundary: Rendering-based training also uses occupancy supervision from input point clouds and geometry/normals preprocessing; it is not supervision from arbitrary photographs alone. Arbitrary topology does not mean artifact-free output, and VAE feed-forward memory is not total generative-pipeline cost.

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

3, 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 · 12 pages