Skip to content
Computational Pathology Reading Roadmap

Writing study: Multi-View 3D Point Tracking

← ICCV 2025 oral writing corpus

Frano Rajič, Haofei Xu, Marko Mihajlovic, Siyuan Li, Irem Demir, Emircan Gündoğdu, Lei Ke, Sergey Prokudin, Marc Pollefeys, Siyu Tang · Official ICCV 2025 oral record

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
≈ 8.05 page-equivalents
Main figures / tables
4 / 3
Attached appendix figures / tables
0 / 0
Attached appendix span
No boundary detected; not assumed zero
Before references / appendix: ≈ 80.5%References / intervening material: ≈ 19.5%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 headingStarts on PDF pageEstimated spanShare of entire PDF
1. Introduction1≈ 0.73 pp.7.3%
2. Related Work2≈ 0.52 pp.5.2%
3. Method3≈ 2.1 pp.21%
4. Experiments5≈ 3.09 pp.30.9%
5. Discussion and Future Work8≈ 0.38 pp.3.8%
6. Conclusion8≈ 0.11 pp.1.1%
Acknowledgements8≈ 0.19 pp.1.9%
References9≈ 1.95 pp.19.5%
All detected figure and table locations

Main text

Attached appendix

    No recognized caption labels.

    Argument and organization

    1. 1 Introduction; 2 Related Work

      Separates monocular depth ambiguity, short-range scene flow and expensive multi-camera optimization.

    2. 3 Method

      Defines synchronized calibrated inputs, a fused feature cloud, directed local correlations, iterative tracking and sliding-window supervision.

    3. 4 Experiments; 5 Discussion; 6 Conclusion

      Evaluates depth-specific datasets, representation controls, number of views and camera layouts before discussing depth failure and scene normalization.

    Problem, claim and contribution

    The claim is a learned multi-view 3D tracker with practical camera counts, not unrestricted 3D tracking from arbitrary video.

    Abstract and Section 1, physical pp.1–2 · Main paper

    Argument and section order

    The input contract precedes point-cloud encoding, local correlation, transformer updates and window propagation, keeping geometry and temporal bookkeeping separate.

    Sections 3.1–3.6, pp.3–5 · Main paper

    Figures, captions and visual explanation

    The overview follows an actual track through feature lifting, a neighborhood inset and an iterative-update loop; blue blocks explicitly denote trainable models.

    Figure 2, p.4 · Main paper

    Tables, comparisons and evidence

    The table gives training sources, feed-forward/online and multi-view flags, metric directions and distinct depth sources; the qualitative grid uses held-out camera viewpoints and empty circles for occluded predictions.

    Table 1 and Figure 3, p.6 · Main paper

    Appendix and reproducibility

    The main text routes depth quality to Appendix A, timing to B, ablations to C, baseline adaptations to D and metric definitions to F; these pointers alone were inspected.

    Sections 3–5, pp.3,5,7–8 · Main paper

    Clarity and writing conventions

    The discussion candidly identifies inaccurate sparse-view depth, bounded overlapping views and manually or heuristically chosen normalization as limits.

    Section 5, p.8 · Main paper

    What to reuse—and what not to infer

    Transferable lesson: Place input privileges and baseline adaptations directly in captions, and distinguish sensor depth, estimated depth and optimization-derived depth.

    Boundary: The reported 7.2 FPS is conditional on supplied sensor depth and is not a complete estimated-depth pipeline timing. Synthetic training and bounded scene normalization limit the transfer claim.

    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, 6

    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.