Kinetics challenge announcements: reminder to make one entry public + send method description

[Joao Carreira]

Dear participants in the Kinetics challenge,

please do not forget to make one of your leaderboard entries public if you want to have a chance to be nominated as winner. 

You must also send us a description of your method and answer the “entry questions” attached below. Deadline for this is tomorrow (11th of June) at 14:00pm GMT.  We will decide winners then and invite the top entries to send us a recording with a talk ahead of the workshop.

We have also update the challenge website with this information.

Please send the descriptions and responses to lsmaira@google.com

Best regards,

The 2021 Kinetics challenge organizers

---- Entry questions  ----

Please answer the following questions to give details about your model and making sure you have followed the rules of the competition

* For the supervised track:

- Have you used an ImageNet checkpoint to initialize your model? If yes, provide a description of the checkpoint used.

- Is your model using audio to compute the representations?

- Is your model using flow to compute the representations? If yes, provide a clear reference to the flow extract method used.

- [RULE CHECK] Can you confirm that you have not used any source of extra supervision other than ImageNet and the labels of the train+valid set of Kinetics-700 2020 dataset?

* For the self supervised track:

- [RULE CHECK] Can you confirm that you only used the videos of the train+val of the seen classes of the Kinetics-700 2020 dataset without the labels? In particular you should confirm that you have not used any other datasets to train your model.

- Is your model using audio to either train the representations or extract the representations?

- Is your model using flow to either train or compute the representations? If yes, provide a clear reference to the flow extract method used.