The allowed user interaction of the algorithms assessed: 

•  The submitted algorithm should automatically perform action recognition and/or action anticipation without any user intervention.

The usage of training data: 

•  The guidelines for the usage of our dataset are adapted from the end user agreement of ImageCLEFmedVQA 2020 [11]. The additional dataset is allowed for the challenge but will not be eligible for the prize to ensure the fairness of the challenge. The additional dataset should be included in the report of the submission of the model and methodology description.

Members of the organizers' institutes may participate in the challenge and are NOT eligible for awards.

The award policy:

• Three categories of awards correspond to each challenge: Action recognition, Action anticipation, Procedure recognition, and Visual Question Answering.
• Please see the "Prizes" page for more information 

Result Announcement:

• For the tasks in Track 1 and Track 2, the top three performing methods will be announced publicly and awarded with corresponding cash prices. The performance results will be released to promote a better algorithm once the dataset is available publicly
• Participating teams can choose whether the performance results will be made public.

Publication: 

• The participating teams may determine who qualifies as authors. They may publish their results separately once the winners of the challenge are announced. However, they shall wait for the dataset to be released publicly before their publication.