ECCV 2026
28 / 28 inventoried oral papers have inspected writing studies. Snapshot: 2026-10-02.
ECCV — writing guide
Make the visual or physical question concrete, then make the intervention testable. Explain the available observations, the representation or acquisition constraint, what changes, and which experiment can distinguish your explanation from alternatives.
This guide synthesizes the 28 oral papers in the ECCV 2026 official program. The examples range from optical sensing and geometric theory to representation interpretation and streaming systems. They do not establish one mandatory ECCV style or a formula for acceptance. The reviewer guidance emphasizes technical soundness and knowledge contribution; a new state-of-the-art accuracy is not universally necessary.
Choose the level of detail
Section titled “Choose the level of detail”| You want to… | Start here |
|---|---|
| Compare figures, tables, PDF lengths and section allocation | 28-paper corpus atlas |
| Plan a model, objective, theory, benchmark, analysis or system paper | Contribution-specific writing plans |
| See a diagram, result table or plot design directly | Inline visual gallery |
| Learn how claims, mechanisms and evidence connect | Writing patterns |
| Search an oral paper, author or session | All paper studies |
| Check initial versus final formatting | 2026 rules and official template |
Writing in brief
Section titled “Writing in brief”Claims and wording: state the information contract
Section titled “Claims and wording: state the information contract”Define inputs, supervision, physical measurements, external priors and inference-time refinement before using words such as efficient, general or zero-shot. Keep the same scientific nouns in the abstract, contribution list, method figure and conclusion. Separate measured behavior from its proposed explanation.
In CoVT, supervision paths and optional dense inference outputs are distinct. That distinction belongs in the explanation of the contribution, not a footnote after the results. In computational imaging, similarly distinguish a physical acquisition from a reconstruction or a learned prior.
Structure: let the question determine the outline
Section titled “Structure: let the question determine the outline”A useful method-paper sequence is observable failure → representation or mechanism → controlled test → scope. A theoretical paper needs definitions, assumptions, a statement and its consequences. A resource paper must first define the new measurement, output or acquisition process. The type guides provide conditional artifact packages, not a compulsory list of sections.
Provable sensing uses a small graph to make the guarantee understandable. FEEL makes hardware, supervision and transfer evidence serve different purposes. Neither should be forced into the same generic architecture outline.
Visualization: show geometry, state and the changed operation
Section titled “Visualization: show geometry, state and the changed operation”Use an overview to orient the reader and a detail inset only when the operation would otherwise remain opaque. Draw coordinate frames, optical paths, temporal state or feature geometry when those are the scientific objects. Label inherited and new components, training-only signals and deployed inputs. A chain of paragraph-filled boxes is not a substitute for showing computation.
Theory can also be visual: second-order texture gives interpretable factors in a result and connects them to acquisition experiments. Keep the conditions of the result beside the diagram. The gallery shows original planning layouts and attributed numerical examples directly.
Evidence and tables: preserve the comparison conditions
Section titled “Evidence and tables: preserve the comparison conditions”Separate main comparisons, controlled interventions and deployment costs. State resolution, acquisition budget, model/data access, metric direction, units and uncertainty. For qualitative evidence, align scenes and regions and explain selection; a chosen crop does not establish an average effect.
GeoSR distinguishes a trained model’s reliance on geometry at inference from changes to its training. NURBS Splatting retains an ablation with better quality but longer optimization. Those non-wins improve the reader’s understanding of the trade-off.
Supplement: a verification route, not a second hidden method
Section titled “Supplement: a verification route, not a second hidden method”The initial ECCV main paper must stand alone, with supporting appendices submitted separately. Keep the defining method, necessary assumptions and decisive evidence in the main text. Give exact destinations for calibration, proofs, implementation, complete results and material failures. A main-text reference to supplementary material does not mean that material was inspected in this collection.
The atlas measures the main PDFs only for this edition. It does not treat unavailable separate supplements as zero-page supplements or copy ICLR’s attached-appendix accounting.
Format and polish: single-column LNCS, phase-specific length
Section titled “Format and polish: single-column LNCS, phase-specific length”ECCV 2026 uses its official single-column LNCS template. The initial limit is 14 content pages; the later official camera-ready announcement allowed 15, although older template wording still says 14. That phase-specific override is retained in the requirements, not converted into a timeless ECCV rule.
Use numeric LNCS citations, figure captions below and table captions above. Inspect the compiled PDF at ordinary reading size and in grayscale. Do not import CVPR’s two-column main format or infer future submission requirements from accepted manuscripts.
Edition archive
Section titled “Edition archive”Each edition retains its own sources, coverage and requirements. The synthesis above remains tied to its named edition.
Conference dates
Section titled “Conference dates”ECCV submission deadlines, event periods and locations are listed below, independently of the inspected writing editions. Unannounced dates are not estimated.
Dates for ECCV only, newest edition first. Submission dates can fall in the previous calendar year. AoE means Anywhere on Earth (UTC−12); other stated timezones are retained. Reconfirm official instructions before submission or travel.
Compare all conferences on the Research Writing timelines →
ECCV 2028
- Submission
- Not announced
- ConferenceNot announced
- Dates not announced
- ECCV runs in even-numbered years. Dates, deadlines and location for the 2028 edition were not established on the checked official conference and ECVA scheduling pages. No projected dates are plotted.
Checked Oct 21, 2026.
ECCV 2026
- Paper registration
- · 23:00CET
- Full paper
- · 23:00CET
- ConferenceMalmö, Sweden
- Sep 8–12, 2026Malmö Arena and Malmömässan
- Paper registration closed February 26. The homepage corrects a briefly displayed March 2 date; later profile and author-enrollment extensions did not extend paper registration.
- These submission deadlines are 23:00 CET, not AoE. Workshops and tutorials were September 8–9; main conference sessions were September 10–12.
Checked Oct 21, 2026.
Apply it to a draft
Section titled “Apply it to a draft”- Name the scientific object, available information, intervention and supported claim.
- Choose a contribution-specific outline and identify the strongest competing explanation.
- Sketch the actual changed computation or acquisition process before polishing its appearance.
- Prepare a matched comparison and a diagnostic that tests the proposed reason for improvement.
- Preserve unfavorable regimes and give a precise supplementary route for reproduction.
- Audit notation, captions, readable scale, anonymity and the exact phase-specific template.
Use the shared argument worksheet to record those decisions. The aim is a clearer scientific contribution, not an imitation of an oral paper’s surface appearance.