| CPC G06Q 20/202 (2013.01) [G06T 7/70 (2017.01); G06V 20/44 (2022.01); G06T 2207/30196 (2013.01)] | 17 Claims |
|
1. A method comprising:
receiving image data of an individual at a location, wherein the location comprises a plurality of point-of-sale (POS) systems at a retail location;
processing the image data using a first trained neural network, wherein the first trained neural network analyzes aspects of an appearance or a behavior of the individual as captured in the image data to quantify a likelihood of the individual participating in a checkout procedure at a first POS system of the plurality of POS systems;
obtaining a first output from the first trained neural network indicating a quantification of the likelihood of the individual participating in the checkout procedure at the first POS system, wherein obtaining comprises applying a machine learning classification model that outputs a probability score representing a confidence that the individual will initiate the checkout procedure at the first POS system within a defined prediction window;
predicting, by the first trained neural network, that the individual will participate in the checkout procedure based at least in part on a determination that the individual will interact with the first POS system within a threshold period of time;
inputting the image data into a second trained neural network configured to determine characteristics of items associated with the individual;
obtaining, from the second trained neural network, a second output indicating one or more of a quantity of the items, or a size one or more of the items, or a weight of one or more of the items; and
managing, via a resource management system communicatively coupled to the plurality of POS systems via a network, resource allocation across the plurality of POS systems based on the first output from the first trained neural network and the second output from the second trained neural network, wherein managing comprises managing an allocation of processing resources among checkout-related tasks and non-checkout-related tasks assigned to the plurality of POS systems, wherein managing resource allocation comprises at least one of:
assigning a checkout-related task to the first POS system, wherein the checkout-related task comprises at least one of scanning an item, searching for information associated with the item, or weighing the item,
unassigning a non-checkout-related task from the first POS system, wherein the non-checkout-related task comprises running at least a portion of a third trained neural network model configured to facilitate at least one of produce recognition during a checkout procedure or theft detection,
terminating an isolated execution environment instantiated on the first POS system, the isolated execution environment being configured to perform the non-checkout-related task, or
assigning the first POS system to perform a non-checkout-related task based on a determination that the individual is not expected to interact with the first POS system, wherein the resource management system communicates with the plurality of POS systems over a network interface.
|