I noticed that there are '2's in column GLASSES when it's supposed to have only '0' (no glasses) and '1' (glasses). Can this be corrected?
Posted by: suhk @ April 23, 2020, 4:55 p.m.Hi suhk,
you are right, there are images with 2 in glasses column (15 in the whole dataset, 13 in the trainset with these ids: 146886, 147119, 148283, 149770, 152017, 152997, 153002, 153258, 155344, 155346, 155419, 158004, 158635). It remained there from the raw annotations, the annotators were asked to distinguish between transparent glasses (1) and sunglasses/tinted glasses (2) and it looks like it slipped through the conversion scripts (since there are so few of them). You can treat images with glasses=2 as if they were glasses=1, but we will nevertheless fix it.
Tomas