Is transfer learning allowed?

Are we allowed to pre-train on another dataset, like ImageNet, or should the model only use the provided datasets?

Thank you!

Hi, thanks for your interest in the competition. It should be noted that all the submitted models will be trained from scratch on the test datasets for evaluations. Thus pre-training on other datasets will not help in this regard.

Well, does it mean it is not forbidden either?

Hi @rauf_kurbanov @isak_westerlund, sorry for the late response as I had to discuss this with our team. We are allowing transfer learning in terms of pre-trainings. It means that you can use any data to pre-train your models and use those pre-trained weights to initialize your submitted models. I will also put it in a separate post so that more people can see that. Thanks for your questions!


@waleedgondal learning about this competition and representational learning in general… so when I submit my code, the training happens on the test dataset from the beginning, no matter how many iterations we give?

really cool competition. thank you.

just saw the post regarding the compute and time limits… so my doubt is clarified.

1 Like