The model should be trained in a way that it can recognize the SK values or the unique objects in the image (shop image) and the quantity of each unique object. It should measure the dimension of the shop.
I use this repo for a competition. I do not share the model in this repo because of rules and regulations of the competition.