Generation of two-dimensional and three-dimensional images of items for visual recognition in checkout apparatus

Abstract

Methods, systems, and computer programs are presented for capturing color images of items for a machine-learning model. In one method, a request, for adding an item to be visually recognized during sales transactions, is detected at a checkout apparatus. Further, the method includes capturing and storing a color image, by each of one or more cameras of the checkout apparatus, of an examination space when the item is present. Further, the capturing and storing of color images of the item is repeated at different poses until at a predetermined number of color images of the item are captured. Further, the method includes generating, by a processor of the checkout apparatus, additional images of the item by modifying the captured color images. A machine-learning model for item recognition is built by training a machine-learning tool with the captured images and the additional images.