Wrong classes in dataset, just "object"

[Julietty]

Hi!

Have strange problem: all classes in the annotations are always "object", checked both links. Would you please share correct labels?

Thanks

[fransafu]

The "Precise Detection in Densely Packed Scenes" project doesn't classify products by class. That's why they use a generic class to denote the "object". It's a detection problem

[Isalia20]

Yes but in the image(where there is a comparison between datasets) on the main page of the README it states that there are 110,712 classes and 86 classes per image on average, however the classes aren't provided in the annotations which makes this comparison(those 2 columns at least) useless

[tojimahammatov]

The SKU-110K dataset is for detecting objects in dense scenarios, not for recognizing what they are actually. The idea is to extract an item in the image whether it is choco, cola, fanta etc. The dataset is built & labelled for that purpose.