About the ModaNet Fashion Understanding Challenge category

Discussions related to Modanet Fashion Understanding Challenge 2018

Can you let me know why the segmentation mask is of lengths in different images?

@kris-singh if you are talking about the segmentation ground truth, this follows the COCO format, which you should easily export to PASCAL VOC format mask or something else. e.g. you can find the instance segmentation ground truth visualization examples from other user: https://github.com/eBay/modanet/blob/dd1b46748078c338b29a5563851b11adec7ca546/ModaNet.ipynb.

Thanks. There are still many missing files even with the updated paperdoll dataset.
for eg:

  1. http://images0.chictopia.com/photos/Adiina/3231876445/atmosphere-boots-new-yorker-jacket-random-brand-sunglasses_400.jpg
  2. http://images0.chictopia.com/photos/AndreXoko/4480710068/sky-blue-blouse-aquamarine-t-shirt-yellow-suit_400.jpg
    and many others.
    P.S. I tried with other urls also but it did not work out.

There should be no images missing. You can check Paper doll dataset update:

Challenge only provide mAP performance in COCO’s way. Can we can performance of each class as Modanet paper provided?
I’d like to know which class is poor and good.