MaCVi: common questions

MaCVi is the Maritime Computer Vision initiative: a community connecting research on aerial, surface and underwater perception. It brings together datasets, benchmark challenges, workshops, research papers, invited talks and discussion. See how the initiative started and the wider maritime community.

You can enter an open benchmark challenge, submit research to an active workshop call, attend talks, contribute a dataset or tool, propose a challenge, or join the discussion. Start with current challenges, the community directory, or Discord.

Challenge entries and regular workshop papers are separate routes. Check the chosen challenge’s rules for the required predictions, code or method description. Selected teams may be invited to contribute to a challenge-summary paper; entering a challenge does not automatically establish authorship.

Use the relevant workshop’s paper schedule and challenge schedule. Paper and challenge deadlines can differ. A closed paper call does not mean an ongoing challenge has closed. Check each deadline’s stated time zone.

MaCVi website accounts are used for supported challenge uploads, teams and results. Register through Sign up or Login. Workshop papers are handled separately on OpenReview. Use the account and team instructions on your challenge page; one website login does not create an OpenReview profile.

For workshops using the Teams & paper page, create or join the correct challenge team and confirm each proposed author’s publication details by the stated metadata deadline. An active OpenReview profile may be required. Keep a pending profile marked as pending and contact the organisers if activation is delayed. Final inclusion and author order are confirmed by the organisers.

Follow your challenge page for its file format, permitted inputs, submission allowance, reset time, evaluation phase and final-entry selection. There is no single quota shared by every MaCVi challenge. A format error, incomplete result or deleted upload may be treated differently by different tracks; the published track rules are the reference.

This depends on the challenge. Read its training, pretraining and validation/test restrictions, and disclose external resources as requested. A dataset being downloadable does not imply that every split is permitted for training.

Check the licence distributed with the specific dataset and any component datasets. Access, redistribution and commercial-use terms can differ. Cite the relevant dataset and benchmark papers; the publication references provide a starting point.

Check the address you registered with, your spam folder and the account-recovery options on the login page. For challenge-account issues, use Discord or contact the listed track organisers. OpenReview activation is a separate process: follow its profile guidance and tell the workshop organisers if moderation is still pending.

Contact the organisers through the MaCVi community. A short outline is enough to start: the research question, data and access terms, intended task and metric, available baselines, timeline, and the people who would maintain it. Proposals are discussed with the relevant workshop team.

Browse the workshop archive, CVPR 2026 summary and recordings, dataset examples and leaderboards. For annotation errors or evaluation questions, include the dataset version and relevant public sample or submission ID when contacting the track organisers.