Item type identification for checkout verification

A method, comprising:

• training a machine-learning-model (MLM) on images of items to provide item types for the items, wherein during training, the MLM is provided an actual item type classification that is expected as output for each of the images in a training data set, and through a this training process, the MLM is configured to provide an item type classification based on an item image received during a particular transaction;
• receiving at least one item image during a transaction at a terminal responsive to an operator of the terminal indicating that a transaction item associated with the at least one item image is a produce item;
• obtaining a current item type for the transaction item from the MLM by providing the at least one item image to the MLM as input;
• making a determination to suspend the transaction for an audit of the transaction item based on the current item type not comporting with a produce item type entered for the transaction;
• wherein training further comprises training the MLM on images depicting a plurality of items, a subset of item images are of non-produce items and a subset of item images are of produce items;
• preprocessing the images prior to providing the images to the MLM for training by cropping background image data;
• and
• responsive to the determination associated with suspending the transaction, receiving an override from the terminal, the override provided from an attendant audit that confirms the transaction item is the produce item, and flagging the at least one item image as feedback to retrain the MLM using the at least one item image and an indication that the transaction item comports with the produce item type.

A system, comprising:

• at least one server comprising at least one processor and a non-transitory computer-readable storage medium;
• the non-transitory computer-readable storage medium comprising executable instructions;
• the executable instructions when executed by at least one processor cause the at least one processor to perform operations, comprising:
• receiving one or more images of an item from a scanner of a camera of a terminal, wherein the one or more images identified by an operator of the terminal as a produce item during a transaction at the terminal;
• preprocessing the one or more images by cropping background image data before providing the one or more images as input to a machine-learning model (MLM);
• processing the MLM a with the one or more images and receiving as output from the MLM an item type for the item;
• and
• causing the transaction to be suspended based on the item type outputted by the MLM not comporting with a produce item type associated with the produce item enabling an accurate item type classification relevant to a particular type of retail shrinkage during a self-checkout;
• wherein the MLM is specifically trained to detect the particular type of retail shrinkage caused by customer identification of non-produce items as lower-cost produce items during self-checkouts;
• and
• wherein the MLM is trained on both item images and corresponding price-lookup (PLU) codes to determine whether a particular item depicted in a particular item image corresponds to an operator-provided PLU code.