FRCSyn-onGoing (based on WACV 2024 FRCSyn: Face Recognition Challenge in the Era of Synthetic Data) Forum

Go back to competition Back to thread list Post in this thread

> Data augmentation with generative models

As we can see, there is a new restriction on using generative models for producing more synthetic training data. Can we at least use generative models for "online" data augmentation (within mini-batches, created during the training of a face recognition model)?

Posted by: Versus @ Sept. 18, 2023, 2:18 p.m.

Can we use generative models, trained only on the provided datasets?

Posted by: Versus @ Sept. 18, 2023, 2:26 p.m.

Dear Versus, in FRCSyn we do not allow the use of Generative models to generate more data. Therefore, the answer to your questions is 'no' in both cases. Please be aware that traditional data augmentation techniques (e.g., flip, rotate) which do not involve generative models are allowed.

Posted by: FRCSyn @ Sept. 21, 2023, 1:02 p.m.

Thank you for the clarification.

Posted by: Versus @ Sept. 21, 2023, 1:05 p.m.
Post in this thread