ACM MMSports 2022

5th International ACM Workshop on Multimedia Content Analysis in Sports

Challenges

This year, MMSports proposes a competition where participants will compete over State-Of-The-Art problems applied to real-world sport specific data. The competition is made of 4 individual challenges, each of which is sponsored by Sportradar with a $1,000 prize. Each challenge comes with a toolkit describing the task, the dataset and metrics on which participants will be evaluated.

Instance segmentation

Challenge toolkit
Challenge submissions

Organizer: Maxime Istasse

Player re-identification

Challenge toolkit
Challenge submissions

Organizer: Vladimir Somers

The challenges are hosted on EvalAI where participants will submit the predictions of their model. Labels of the evaluation set used to rank the participants will be kept secret and leaderboards will display the ranking for each challenge.

Datasets

Each challenge is based on a dataset of manually annotated images made publicly available by Sportradar. The proposed solutions must not use any other data than the data specifically given in the challenge toolkit. However, for neural network solutions, using a pretrained model from a well known model-zoo and trained on a public dataset is accepted, but this must be clearly stated in the submission report. The data destinated for evaluation cannot be used in any form for training (pseudo-labelling is forbidden).

Awards

The challenges will run in two phases described hereafter. Participants are allowed (and encoraged) to submit to both phases, and can compete in the 4 different challenges. Any team of one or multiple members can participate in the challenge but individuals can only belong to one single team per challenge. Challenge organizers cannot participate in their challenge.

First phase - Compete to have your method published in the conference proceedings

For each challenge, the top 3 teams performing best on the leaderboard will have a 2 weeks delay to write a workshop paper presenting their method. The papers will follow a reviewing process and, upon acceptance by reviewers, will be published on the workshop proceedings. Please refer to the workshop call for papers for submission policies.

11:59 p.m. AoE, July 4, 2022 First phase closes for all challenges. Winners are notified and should start preparing their paper
11:59 p.m. AoE, July 18, 2022 Challenge papers submission
July 29, 2022 Acceptance notification
11:59 p.m. AoE, August 5, 2022 Camera-ready submission

Second phase - Compete to win up to $1,000

For each challenge, the team performing best on the leaderboard will be awarded a $500 prize. In addition, a jury prize of $500 per challenge will be awarded to the team proposing the most innovative solution, regardless of the performance on the leaderboard.

To be eligible, submissions must be acompagnied with a link to a public repository hosting the code to reproduce their results. The repository should also include a technical report describing the proposed approach. The reports and the code will be reviewed by the challenge organizers that may disqualify any submission for which legitimity is uncertain.

The winners will be contacted one week before the workshop to prepare a 5 minutes presentation presenting their method. Their technical report and a link to their repository will be published on the challenge website.

11:59 p.m. AoE, September 20, 2022 Second phase ends for all challenges. Submissions must include technical report with link to code repository.
October 3rd, 2022 Winners are contacted privately and should start preparing a presentation of their approach
October 10th, 2022 Workshop day: Winners are officially announced

Questions and contacts

Any question related to the challenges should be asked on our discord community where updates and notifications about the challenges will be shared.

The discord community can also be joined using this linkq: https://discord.gg/JvMQgMkpkm.

Want to be in our mailing-list?