Writing study: Knowledge Distillation for Learned Image Compression

ICCV overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← ICCV 2025 oral writing corpus

Yunuo Chen, Zezheng Lyu, Bing He, Ning Cao, Gang Chen, Guo Lu, Wenjun Zhang · 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: Efficiency, inference & 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

11 / 4

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	≈ 0.81 pp.	7.36%
2. Related Work	2	≈ 0.63 pp.	5.73%
3. Method	3	≈ 3.22 pp.	29.27%
4. Experiments	6	≈ 2.19 pp.	19.91%
5. Conclusion	8	≈ 0.31 pp.	2.82%
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. 5
• Figure 4 · p. 5
• Figure 5 · p. 6
• Table 1 · p. 7
• Figure 6 · p. 7
• Figure 7 · p. 7
• Table 2 · p. 7
• Figure 8 · p. 7
• Table 3 · p. 7
• Figure 9 · p. 8
• Figure 10 · p. 8
• Table 4 · p. 8
• Figure 11 · p. 8

Attached appendix

No recognized caption labels.

Argument and organization

1. 1 Introduction; 2 Related Work

   Separates rate-distortion quality from model complexity and identifies architecture mismatch and supervision as distillation barriers.

2. 3 Method

   Defines LIC components, modular transform/entropy distillation, teacher-shaped student construction and implicit end-to-end supervision.

3. 4 Experiments; 5 Conclusion

   Separates the resulting codec, distillation-versus-vanilla controls, architecture alignment and extensions to two other teachers.

Problem, claim and contribution

Distinguishes the general SMoDi training framework from KDIC, a particular compressed S2CFormer-derived student.

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

Argument and section order

The rate-distortion objective precedes modular decomposition; the channel-energy diagnosis then motivates why plain intermediate-feature matching is insufficient.

Sections 3.1–3.4 and Figure 4, pp.3–6 · Main paper

Figures, captions and visual explanation

The overview labels six teacher/student stages, tensor resolutions and channel/block counts, with frozen/trainable symbols exposing the active module and inherited context.

Figure 2, p.4 · Main paper

Tables, comparisons and evidence

Table 2 includes teacher, vanilla student and distilled student, separating architecture savings from training gains; Table 1 states dataset-specific BD-rate alongside parameter, FLOP, latency and training-throughput measures.

Tables 1–2, p.7; timing protocol, pp.6–7 · Main paper

Appendix and reproducibility

The preliminary Jacobian/KL rationale explicitly defers detailed proof and discussion to the appendix; only the main argument and pointer were scanned, not proof-verified.

Section 3.2.1, p.4 · Main paper

Clarity and writing conventions

The extension section calls its two added student designs examples rather than an exhaustive search, appropriately narrowing the generalizability claim.

Section 4.5, p.8 · Main paper

What to reuse—and what not to infer

Transferable lesson: Always retain teacher, un-distilled student and distilled student rows so architecture compression and supervision benefits cannot be confused.

Boundary: Figure 1 describes a 2.5% BD-rate reduction versus the teacher, but Table 2 attributes the 2.50-point improvement to the vanilla student; the teacher remains better in BD-rate. The complexity and throughput protocols use different image sizes and should not be combined into one speed 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, 7, 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