Process of recognition and grouping of objects into preset categories
Annotation of objects in photos to train systems to recognize and interpret them
the sale of goods and services to consumers
Process of locating instances of objects with Bounding Box
The dataset collected using crowdsourcing contains images of products of different categories on supermarket shelves. The goods in the photos are placed in an arbitrary position: with the cover facing the buyer and away from the buyer, sideways and at an angle. 1/3 of the photos have empty and half-empty shelves.
For all images, there is a marking with polygons of each visible product.
The detection of goods and classification by categories and types of placement was carried out in CVAT. The data is presented in the form of source photographs, masks with annotations, and attribute coordinates in an XML file.