Skip to content
Computational Pathology Reading Roadmap

Writing study: Towards a Unified Copernicus Foundation Model for Earth Vision

← ICCV 2025 oral writing corpus

Yi Wang, Zhitong Xiong, Chenying Liu, Adam J. Stewart, Thomas Dujardin, Nikolaos Ioannis Bountos, Angelos Zavras, Franziska Gerken, Ioannis Papoutsis, Laura Leal-Taixé, Xiao Xiang Zhu · Official ICCV 2025 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
12 physical pages
Before references / appendix
≈ 8.16 page-equivalents
Main figures / tables
5 / 5
Attached appendix figures / tables
0 / 0
Attached appendix span
No boundary detected; not assumed zero
Before references / appendix: ≈ 68%References / intervening material: ≈ 32%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≈ 0.96 pp.8%
2. Related work2≈ 0.52 pp.4.33%
3. Copernicus-Pretrain2≈ 1.46 pp.12.17%
4. Copernicus-FM4≈ 2.54 pp.21.17%
5. Copernicus-Bench6≈ 1.38 pp.11.5%
6. Bridging EO & climate via grid embeddings8≈ 0.56 pp.4.67%
7. Conclusion8≈ 0.27 pp.2.25%
Acknowledgment9≈ 0.12 pp.1%
References9≈ 3.84 pp.32%
All detected figure and table locations

Main text

Attached appendix

    No recognized caption labels.

    Argument and organization

    1. 1 Introduction; 2 Related work

      Maps sensor diversity, model flexibility and evaluation breadth to three coordinated contributions.

    2. 3 Copernicus-Pretrain; 4 Copernicus-FM

      Documents global grid sampling and missing modalities before dynamic patchification, metadata and training objectives.

    3. 5 Copernicus-Bench; 6 EO and climate; 7 Conclusion

      Introduces a licensed task inventory and common protocols, then separates a modest climate regression demonstration from future forecasting.

    Problem, claim and contribution

    The data, model and benchmark claims each answer one named bottleneck, making the three-part contribution internally traceable.

    Sections 1 and 7, physical pp.1–2,8 · Main paper

    Argument and section order

    Sampling and curation precede architecture; a benchmark dataset table appears before results, so sensor coverage and evaluation tasks are explicit.

    Sections 3–5, pp.2–7 · Main paper

    Figures, captions and visual explanation

    The pipeline separates spectral/variable hypernetworks, metadata encoding and frozen distillation teachers; reconstruction paths are drawn separately for eight sampled modalities.

    Figure 4, p.5 · Main paper

    Tables, comparisons and evidence

    The benchmark table preserves task-specific metrics and unavailable modality entries, reports three-run means and deviations, and distinguishes frozen encoders from supervised training.

    Tables 3–4 and Section 5.2, p.7 · Main paper

    Appendix and reproducibility

    Dataset analyses, subtractive/metadata ablations, implementation, new-task curation and global embedding details are assigned to the appendix; only those pointers were inspected.

    Sections 3.2–6, pp.4,6–8 · Main paper

    Clarity and writing conventions

    The closing limitation explicitly restricts scope to Sentinel sensors and roughly one year and places native multimodal fusion/time series support in future work.

    Section 7, p.8 · Main paper

    What to reuse—and what not to infer

    Transferable lesson: For a data–model–benchmark package, give each deliverable a separate contract and preserve unsupported task cells instead of manufacturing a universal average.

    Boundary: The 18.7M image count is not a count of independent global sites; city-centered sampling and missing modalities shape coverage. Ten-year climate-summary regression is not demonstrated weather forecasting, and arbitrary-sensor architectural support is broader than the validated sensor scope.

    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
    5, 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.