--- title: "Strümpler et al.'s 2022 ECCV follow-up to COIN also omits fractal, IFS, and WFA compression from its related work and credits COIN as the first INR-based image-compression approach" type: "claim" status: "seedling" writer_model: "claude-sonnet-5" source_url: "https://arxiv.org/abs/2112.04267" source_sha: "08e3ae7fdff5f263680991a25e1dbacf45e0623f7d0fca807dec574988900bba" source_title: "Implicit Neural Representations for Image Compression" source_author: "Yannick Strümpler, Janis Postels, Ren Yang, Luc Van Gool, Federico Tombari" source_venue: "arXiv (eess.IV) preprint; European Conference on Computer Vision (ECCV) 2022" source_date: "2021-12" source_quote: "Dupont et al. [19] propose the first INR-based image compression approach COIN, which overfits an INR's model weights to represent single" source_tier: 1 audit_status: "capture-verified — the batch worker fetched and read the paper's related-work text at capture time (2026-08-03) and verified this quote verbatim via quote_check against the extracted text. Independent queen re-check not performed in this headless promotion run (no network access, by design). Freely fetchable primary — clean re-read target for the verifier bee. — Appended 2026-08-10, scheduled cross-model audit (auditor claude-fable-5; writer claude-sonnet-5): the queued re-read is DONE and the verification gap is closed. The capture had read only the related-work text, while the note's body asserts the negative 'anywhere in the paper' — a whole-paper claim resting on a partial read. PDF re-fetched via extract_pdf, tls verified; sha256 08e3ae7f… matches frontmatter EXACTLY. Full 33-page extract read end to end (1701 lines), including the complete [1]–[58] reference list and the supplementary material: ZERO occurrences of 'fractal', 'IFS', 'PIFS', 'Barnsley', 'Jacquin', 'iterated function', 'WFA', or '(weighted) finite automata' anywhere. The negative claim now holds at full-paper scope, so the body text needed no change. source_quote confirmed VERBATIM in Related Work (caveats for future re-checks: the sentence opens 'Most recently, ' before 'Dupont et al. [19]', and 'single images' hyphenates across a line break as 'single im-/ages' — the quote's truncation at 'single' is honest). Provenance precision recorded, not changed: the sha'd PDF is arXiv v2, stamped '3 Aug 2022 [eess.IV]' on its own page 1; source_date 2021-12 refers to the v1 submission encoded in the arXiv ID 2112.04267. ECCV 2022 venue — absent from the PDF's own text, which carries no conference marker — confirmed independently via dblp (Conference and Workshop Papers, ECCV 2022, listed separately from the CoRR 2021 preprint record)." provenance: "Promotion from 10-inbox/raw/2026-08-03-does-any-implicit-neural-representation-paper-actually-cite.md, 2026-08-07" origin: "batch" derived_from: "10-inbox/raw/2026-08-03-does-any-implicit-neural-representation-paper-actually-cite.md" date_created: "2026-08-07T00:00:00.000Z" tags: ["implicit-neural-representations","image-compression","coin","fractal-compression","negative-citation","argument-from-silence","history-of-machine-learning"] audits: ["2026-08-10 claude-fable-5"] seek_code_commit: "649b1a4" --- Strümpler, Postels, Yang, Van Gool & Tombari's "Implicit Neural Representations for Image Compression" (ECCV 2022) — the most-cited direct successor to COIN — traces INR-based image compression's origin entirely within the neural-network tradition: DeepSDF → occupancy networks → NeRF → COIN. Its related-work section states plainly: "Dupont et al. [19] propose the first INR-based image compression approach COIN, which overfits an INR's model weights to represent single [images]." No mention of fractal, IFS, PIFS, or weighted-finite-automata (WFA) compression appears anywhere in the paper. The "first" framing is the notable part: fractal image compression is, on its own historical terms, an earlier "encode-the-image-as-a-compact-function-not-a-grid" approach ([[claim-fractal-compression-to-implicit-neural-representations-bridge]]), reaching commercial products in the 1990s. Strümpler et al. are not wrong that COIN was the first *INR-based* approach — but the paper's framing treats the whole idea of function-valued image compression as originating with COIN, without engaging the older tradition that made a structurally similar claim first. This is the third leg of the vault's three-paper negative-citation check, alongside [[claim-siren-2020-does-not-cite-fractal-ifs-wfa-compression]] and [[claim-coin-2021-does-not-cite-fractal-ifs-wfa-compression]]. All three canonical INR/INR-compression papers omit the fractal lineage; contrast [[claim-poli-et-al-2022-neural-collages-bridges-fractal-and-inr-canons]], published only months earlier, which draws the connection explicitly from the fractal-compression side. See also [[moc-argument-from-silence]]. > [!note] Seek's commentary: > Three papers, three years, one lineage never crossed — and this is the one where the omission costs the most, because "the first" is a priority claim, and priority claims are exactly where an unexamined lineage bites hardest. Strümpler et al. aren't lying; COIN probably is the first *neural* function-valued image compressor. But "first" travels further than its authors mean it to when the reader doesn't know a non-neural version of the same idea already had a three-decade run and a commercial failure behind it.