Writing study: OpenDance: Multimodal Controllable 3D Dance Generation with Large-scale Internet Data
← CVPR 2026 oral writing corpus
Jinlu Zhang, Zixi Kang, Libin Liu, Jianlong Chang, Qi Tian, Feng Gao, Yizhou Wang · Official CVPR 2026 oral record
Figure, table and section ledger
Editorial contribution type: Datasets, benchmarks & measurement (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
- ≈ 8.08 page-equivalents
- Main figures / tables
- 5 / 9
- Attached appendix figures / tables
- 0 / 0
- Attached appendix span
- Separate file recorded below; not in this PDF total
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 | ≈ 0.88 pp. | 8% |
| 2. Related Works | 2 | ≈ 0.48 pp. | 4.36% |
| 3. OpenDanceSet Dataset | 3 | ≈ 1.63 pp. | 14.82% |
| 4. OpenDanceNet | 4 | ≈ 2.47 pp. | 22.45% |
| 5. Experiments | 7 | ≈ 1.46 pp. | 13.27% |
| 6. Limitation and Future Work | 8 | ≈ 0.07 pp. | 0.64% |
| 7. Conclusion | 8 | ≈ 0.22 pp. | 2% |
| References | 9 | ≈ 2.92 pp. | 26.55% |
All detected figure and table locations
Main text
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 8 physical pages; 7 detected figures and 6 tables. These may include overlapping main-text material.
Located appendix only: starts on physical p.1; ≈ 7.96 page-equivalents, 7 figure labels and 6 table labels. This interval is distinct from the whole file.
Argument and organization
- Dataset before model
Describe collection, annotations and physical refinement before the controllable generator.
- Conditions and refinement
Separate tokenization, joint prediction, spatial supervision and inference refinement before control-specific ablations.
Problem, claim and contribution
The opening frames the missing resource as richly annotated multimodal control, not just music-conditioned motion quality.
Abstract and Introduction · Main paper · Additional source 2
Argument and section order
The dataset pipeline is a full section before OpenDanceNet, making the data and method contributions separately assessable.
Sections 3–4 · Main paper · Additional source 2
Figures, captions and visual explanation
Figure 2 pairs the annotation/processing pipeline with a genre distribution; Figure 5 aligns rendered motion sequences under distinct keypoint and trajectory conditions.
Figure 2, PDF p. 4; Figure 5, p. 7 · Main paper · Additional source 2
Tables, comparisons and evidence
The main comparisons report physical contact, distributional quality, diversity and beat alignment together; control types and refinement steps are tested in separate ablation tables.
Tables 2–9, PDF pp. 7–8 · Main paper · Additional source 2
Appendix and reproducibility
The supplement pairs preference results with confidence intervals and documents a low-height trajectory failure; static sequences do not replace the separate uninspected videos.
Supplement Table 11; §F and Figure 12 · Main paper · Additional source 2
Clarity and writing conventions
Data-cleaning and post-optimization are described before success metrics, preventing the processed internet motions from being presented as direct motion-capture ground truth.
Section 3.4, PDF p. 4 · Main paper · Additional source 2
What to reuse—and what not to infer
Transferable lesson: For generation datasets, connect collection/cleaning to a model that exploits the new controls, and report controllability separately from realism, diversity and physical plausibility.
Boundary: Static dance frames are inadequate evidence of temporal smoothness; distinguish reconstructed internet motion, refinement and original motion-capture references in both captions and claims.
Apply this lesson using the argument and evidence worksheet.
Inspection record and sources
- Checked
- 2026-09-26
- Reading status
- Writing-focused manuscript scan + selected visuals inspected
- Main text
- Manuscript-wide writing scan of main PDF pp. 1–8: framing, section structure, method exposition, figure/table captions, evaluation organization and conclusions; not an exhaustive technical proof audit or reproduction.
- Appendix / supplement
- Scanned the eight-page official ZIP-member PDF for architecture/training settings, metric definitions, codebook and sampling tests, user studies with confidence intervals, annotation details, controllability examples, and failure cases; videos were not viewed.
- Visually inspected PDF pages
- 4, 7
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.