Missing data.py and question about train/val splits

[hongnianwang]

Great work!

I'm trying to use it but ran into two quick things:

1. data.py is missing from the repo.
2. I downloaded the full 2227 images, but unlike the sample folder, there's no train/val split. Is there an official split file, or should we split randomly?

Thanks!

[vigsam-coder]

Hi! Thanks for your interest in our work.

1. We didn’t include data.py because there isn’t an API to download the data from Figshare, so it’s meant to be prepared manually.
   https://figshare.com/articles/dataset/_b_CoatingVision_A_Defect_Dataset_for_Coating_Manufacturing_b_/29260121?file=55182476

2. Regarding the dataset split: there’s no official train/val split provided. You can safely split the 2,227 images randomly, for example using an 80% train / 20% validation split, which works well for most experiments.
   Hope that helps!