Hello.
As we can see, the provided datasets contain images in different formats - some in the form of detected and cropped faces, and others - as full images of people. To use the faces from these datasets for face recognition purposes, we need to pass them through face detector. It is not clear from the "Restrictions" section whether we can use a pre-trained face detector. Please clarify this point.
Dear Versus, thank you for your post. We want to clarify that there are no restrictions on the choice of face detection methods for use in the competition.
Posted by: FRCSyn @ Sept. 21, 2023, 1 p.m.Thank you
Posted by: Versus @ Sept. 21, 2023, 1:04 p.m.Hi, the BUPT-balanced Face folder has rec file for training and testing; however the images in the folder are not aligned and are of arbitrary size. I am confused whether the .rec files are made after cropping and alignment or we may need to do it ourselves and then again make the .rec files?
Posted by: SUVIDHA @ Oct. 17, 2023, 6 a.m.Dear Suvidha,
In the welcome email we suggest to use images at path: BUPT-Balancedface/images/Equalizedface.tar.gz
Participants are free to do cropping and alignment according to their preferences.
Posted by: FRCSyn @ Oct. 19, 2023, 8:11 a.m.