A user is prompted to begin a transaction by presenting a welcome message on a display. Output images from a set of cameras are forwarded to a machine learning model trained to identify items on a checkout tray during the transaction once no movement is detected on the checkout tray. An identification of an error state designating a particular error is received from the machine learning model and, in response, the user is provided with an error message on the display identifying the particular error. Once it is determined that the error state has been cleared, output images from the set of cameras are forwarded to the machine learning model trained to identify the items on the checkout tray. An identification of the items on the checkout tray is received and the identified items are added to a list of items to be purchased shown on the display.