International Conference On Computer Vision

IEEE/CVF International Conference on Computer Vision (ICCV)

Conference positioning

IEEE/CVF International Conference on Computer Vision (ICCV) is a top computer-science conference venue for computer vision across recognition, geometry, video, learning, datasets, and foundation models. It rewards a vision paper with international flagship-level novelty and evaluation breadth. Treat this skill as a fit / venue-selection / re-framing tool for conference submission strategy, not as a substitute for the current year's CFP, author kit, ethics policy, or submission portal.

Because CS conferences change deadlines, templates, page limits, review workflow, artifact rules, AI-use policy, and rebuttal formats every cycle, always verify the live official instructions before making a submission-ready recommendation. Start from the official source anchor recorded for this venue in ../../resources/conference-roster.md and ../../resources/official-source-map.md.

When to trigger

• The author names ICCV / IEEE/CVF International Conference on Computer Vision as the target venue.
• A manuscript in computer vision across recognition needs a conference-fit read before being formatted or submitted.
• The paper must be re-framed from journal style or arXiv style into a selective CS conference narrative.
• The author needs an evidence-gap, anonymity, artifact, rebuttal, or re-routing diagnosis for this venue.

Scope & topic fit

• Core fit: computer vision across recognition, geometry, video, learning, datasets, and foundation models.
• Best submissions make a precise contribution type visible: algorithm, theorem, system, dataset, benchmark, empirical finding, design artifact, tool, or socio-technical analysis.
• The paper should explain why the result matters to ICCV's reviewers, not just why it is interesting to the authors' lab or product context.
• Position related work against the most recent conference-cycle papers in this venue and its closest siblings; stale comparisons are a common early-review weakness.
• If the contribution is interdisciplinary, state which part is CS research and which part is domain evidence.

Venue-specific calibration

• Reviewer lens: Read reviewers as CV specialists. Strong ablations, dataset protocols, qualitative failures, and current vision baselines are mandatory.
• Contribution hook to foreground: the venue-specific contribution bar.
• Scope vocabulary to use naturally in the abstract and introduction: computer vision across recognition, geometry, video, learning, datasets, and foundation models.
• Distinctive fingerprint for reviewer calibration: vision, across, recognition, geometry, video, learning, datasets, foundation, models, venue-specific, contribution, flagship, iccv, thecvf.
• Official anchor domain: iccv.thecvf.com. Quote annual rules only after opening that source and the current-year CFP/author kit.

Close-neighbor routing guardrail

• Route to ICCV when the paper is a computer-vision contribution with flagship-level novelty, strong benchmarks, ablations, and community relevance.
• Compare CVPR and ECCV by current cycle and area fit, not acronym familiarity. Dataset-only, application-only, or medical-imaging claims may fit narrower venues better.

What distinguishes this venue from its closest siblings

• What ICCV is. The IEEE/CVF International Conference on Computer Vision, held biennially in odd years.
• vs ECCV. ECCV is the even-year European biennial counterpart (Springer); with annual CVPR they form the vision big-three.
• Routing. WACV is applied/winter, ACCV is Asia-Pacific regional; ICCV is a top general-vision flagship.

ICCV-specific routing detail

• Prefer ICCV when the paper targets the international computer-vision flagship cycle with strong vision novelty, benchmarks, analysis, and broad vision-community relevance.
• Route European-cycle fit to ECCV, annual North American cycle fit to CVPR, and graphics or visualization claims to SIGGRAPH/EuroVis-family venues when vision is not the main object.
• ICCV evidence should make benchmark protocol, method novelty, ablations, robustness, qualitative analysis, and limits of generalization visible.

Method & evidence bar

• Use current vision baselines, strong ablations, dataset-specific protocols, and qualitative examples that reveal failure modes.
• Keep the anonymous submission self-contained; external material should follow the current-cycle policy exactly.
• For generated or foundation-model outputs, show robustness, data provenance, and evaluation beyond cherry-picked visuals.
• For ICCV, the evidence must support the venue-specific signature: a vision paper with international flagship-level novelty and evaluation breadth.
• Include limitations, negative results, compute/resource reporting, data provenance, and ethics details when they affect the claim.

Structure & house style

• Lead with the visual problem and the technical insight; then prove it across datasets, metrics, and ablations.
• Make figures do work: pipeline, qualitative wins/failures, and compact quantitative comparisons.
• Use the current official template exactly; do not guess page limits, font sizes, supplement rules, anonymity exceptions, or camera-ready requirements from old cycles.
• The introduction should answer: problem, why now, what is new, why this venue, and what evidence proves the claim.
• Put the strongest result in the main paper, not only in the appendix or supplement; reviewers should not have to reconstruct the contribution.

Official-cycle checklist

• Open the live official venue page: https://iccv.thecvf.com/
• Re-check the current cycle's CFP, author kit, submission system, abstract/paper deadlines, page limits, supplementary-material rules, anonymity policy, dual-submission policy, ethics policy, AI-use policy, artifact/code/data expectations, rebuttal/author-response format, and camera-ready requirements.
• Confirm the review workflow and portal: the current CVF/ECCV/CMT/OpenReview author kit and anonymity policy.
• Check whether accepted papers require in-person presentation, separate registration, artifact badges, proceedings copyright, or post-acceptance release forms.
• If the live official instructions conflict with this skill, the official instructions win.

Pre-submission self-check

• One sentence states why this manuscript belongs at ICCV, using the venue's scope rather than generic "top conference" language.
• The claim is calibrated to the evidence: no broader than the datasets, proofs, systems, user studies, deployments, or threat model support.
• Related work includes the nearest current-cycle computer vision flagship papers and explains the technical delta.
• The paper satisfies the current official template, anonymity, ethics, artifact, and rebuttal requirements.
• The main paper is self-contained enough for reviewers to evaluate novelty and correctness without hunting through external links.

Common desk-reject triggers

• A thin architecture tweak with marginal gains and no analysis.
• Using non-comparable baselines, private data splits, or hidden external links that violate review policy.
• Ethics, consent, or biometric/medical claims handled as boilerplate rather than as real constraints.
• Formatting, anonymity, dual-submission, external-link, or supplement violations under the current-year policy.
• A contribution framed for a neighboring field while giving ICCV reviewers too little technical or empirical substance.

Re-routing decision

If the paper misses ICCV's bar, compare against computer-vision-and-pattern-recognition / european-conference-on-computer-vision / winter-conference-on-applications-of-computer-vision / asian-conference-on-computer-vision. Re-route based on contribution type, not prestige: theory to a theory venue, systems to a systems venue, application-heavy work to a domain venue, and early ideas to workshops or shorter tracks when the official CFP supports them.

Output format

[Fit] High / Medium / Low (one-line reason)
[Target] IEEE/CVF International Conference on Computer Vision (ICCV)
[Contribution type] algorithm / theory / system / dataset / benchmark / empirical / design / security / other
[Main evidence gap] <single most important missing proof, experiment, study, artifact, or policy check>
[Official items to re-check] CFP / author kit / deadline / format / anonymity / ethics / AI-use / artifact / rebuttal / camera-ready
[Top rejection risk] <venue-specific risk>
[Re-route suggestion] <better-matched conference or journal if not a fit>

Source: brycewang-stanford/Awesome-Journal-Skills → Computer-Science-Conference-Skills/skills/international-conference-on-computer-vision/SKILL.md