Can I ask whether the evaluation criteria is performed on the eye region only or on the whole image (background included)? From the formula given, it seems the criteria takes the whole image into account. But it seems that synthesizing similar background to ground truth should not be the focus in my opinion.
Hi, actually it’s a very good question. However currently, the criterion evaluates the whole image.
Hello, I made one submission yesterday, and the result is 41.688 which ranks the third. But according to the evaluation criterion, the result should be the lower the better, right?
The server was showing the order of the results in inverted order. It has been fixed please check now.