> The input size

Is there any restriction on the input size?
The benchmark is trained on 224x224.
Can we use 448x448 to train and submit?

Posted by: ust-gaze @ March 5, 2021, 3:44 a.m.

Hi,

The baseline is trained and tested on 224*224. You can train the model with a higher resolution but please test it on 224 sizes. Please also note that you must release the source code if you win the competition. We will check every detail to make sure a fair competition.

Have fun!

Kind regards,
Xucong

Posted by: Xucong_Zhang @ March 6, 2021, 1:07 p.m.

I see. Thanks.

Posted by: ust-gaze @ March 6, 2021, 1:10 p.m.

Hi,

Can we train our models using lesser size of input images ?

Are there any other restrictions on the type of inputs that we can use. Specifically, apart from face images, can we also use eye images in parallel?)

Thanks.

Murthy.

Posted by: raghu_cpdmlrd @ May 4, 2021, 5:21 a.m.

Hi Murthy,

You can also use the data with lesser size, including using the eye image as part of the input.

The principle as well as our motivation for this competition is encouraging novel and interesting solutions for the gaze estimation task. Therefore, as long as you are working on the method which can benefit the community is acceptable.

Cheers,
Xucong

Posted by: Xucong_Zhang @ May 4, 2021, 7:40 a.m.
Post in this thread