Granted Patent B2

US 12,488,587 · App. 18/764,517 · Granted Dec 2, 2025

Inventors: Sumedh Vilas Datar (Grapevine, TX); Tejas Pradip Rode (Coppell, TX); Sailesh Bharathwaaj Krishnamurthy (Irving, TX); Crystal Maung (Dallas, TX)

Assignee: 7-ELEVEN, INC.

G06V20/41G06F18/2148G06T7/55G06T11/20G06T2210/12G06V20/44

Abstract

Claims (69)

1 . A system for capturing images for training an item identification model comprising:

a plurality of cameras, wherein each camera is configured to capture images of at least a portion of a platform;

a memory, operable to store an item identification model, wherein the item identification model is configured to identify items based at least in part upon images of the items;

a processor, operably coupled with the memory, and configured to:

obtain an identifier associated with an item;

detect a triggering event at the platform, wherein the triggering event corresponds to a user placing the item on the platform;

cause at least one camera from among the plurality of cameras to capture an image of the item;

extract a set of features associated with the item from the image, wherein each feature corresponds to a physical attribute of the item;

associate the item to the identifier and the set of features; and

add a new entry to a training dataset of the item identification model, wherein the new entry represents the item labeled with at least one of the identifier and the set of features; and

a three-dimensional (3D) sensor positioned above the platform, wherein the 3D sensor is configured to capture overhead depth images of the item placed on the platform, wherein each overhead depth image is configured to capture upward-facing surfaces of the item placed on the platform;

wherein the processor is further configured to:

cause the 3D sensor to capture a depth image of the item;

determine an orientation of the item with respect to the platform;

determine that the orientation of the item is longitudinal with respect to the platform; and

in response to determining that the orientation of the item is longitudinal with respect to the platform, cause a first subset of cameras from among the plurality of cameras to take one or more images of the item, wherein the first subset of cameras are positioned above the platform.

2 . The system of claim 1 , wherein the set of features comprises at least one of:

one or more dominant colors of the item, wherein each of the one or more dominant colors is determined based at least in part upon a set of pixel colors associated with the item from the image;

a dimension of the item, wherein the dimension comprises a width, a length, and a height of the item;

a bounding box around the item; and

a mask that defines a contour around the item.

3 . The system of claim 1 , further comprising a weight sensor configured to measure weights for items on the platform;

wherein the processor is further configured to:

receive a plurality of weights of multiple instances of the item;

determine a mean of the plurality of weights;

associate the mean of the plurality of weights to the item; and

add the mean of the plurality of weights to the new entry.

4 . A method for capturing images for training an item identification model comprising:

obtaining an identifier associated with an item;

detecting a triggering event at a platform, wherein the triggering event corresponds to a user placing the item on the platform;

causing at least one camera from among a plurality of cameras to capture an image of the item;

extracting a set of features associated with the item from the image, wherein each feature corresponds to a physical attribute of the item;

associating the item to the identifier and the set of features;

adding a new entry to a training dataset of the item identification model, wherein the new entry represents the item labeled with at least one of the identifier and the set of features;

causing a 3D sensor to capture a depth image of the item;

determining an orientation of the item with respect to the platform;

determining that the orientation of the item is longitudinal with respect to the platform; and

in response to determining that the orientation of the item is longitudinal with respect to the platform, causing a first subset of cameras from among the plurality of cameras to take one or more images of the item, wherein the first subset of cameras are positioned above the platform.

5 . The method of claim 4 , wherein the set of features comprises at least one of:

one or more dominant colors of the item, wherein each of the one or more dominant colors is determined based at least in part upon a set of pixel colors associated with the item from the image;

a dimension of the item, wherein the dimension comprises a width, a length, and a height of the item;

a bounding box around the item; and

a mask that defines a contour around the item.

6 . The method of claim 4 , further comprising:

receiving a plurality of weights of multiple instances of the item;

determining a mean of the plurality of weights;

associating the mean of the plurality of weights to the item; and

adding the mean of the plurality of weights to the new entry.

7 . A non-transitory computer-readable medium storing instructions that when executed by a processor cause the processor to:

obtain an identifier associated with an item;

detect a triggering event at a platform, wherein the triggering event corresponds to a user placing the item on the platform;

cause at least one camera from among a plurality of cameras to capture an image of the item;

extract a set of features associated with the item from the image, wherein each feature corresponds to a physical attribute of the item;

associate the item to the identifier and the set of features;

add a new entry to a training dataset of the item identification model, wherein the new entry represents the item labeled with at least one of the identifier and the set of features;

cause a 3D sensor to capture a depth image of the item;

determine an orientation of the item with respect to the platform;

determine that the orientation of the item is longitudinal with respect to the platform; and

in response to determining that the orientation of the item is longitudinal with respect to the platform, cause a first subset of cameras from among the plurality of cameras to take one or more images of the item, wherein the first subset of cameras are positioned above the platform.

8 . The non-transitory computer-readable medium of claim 7 , wherein the set of features comprises at least one of:

one or more dominant colors of the item, wherein each of the one or more dominant colors is determined based at least in part upon a set of pixel colors associated with the item from the image;

a dimension of the item, wherein the dimension comprises a width, a length, and a height of the item;

a bounding box around the item; and

a mask that defines a contour around the item.

9 . The non-transitory computer-readable medium of claim 7 , wherein the instructions further cause the processor to:

receive a plurality of weights of multiple instances of the item;

determine a mean of the plurality of weights;

associate the mean of the plurality of weights to the item; and

add the mean of the plurality of weights to the new entry.

Assignments (1)

Continuity (4)

References Cited (43)