System and method for synthetic image training of a machine learning model associated with a casino table game monitoring system

Patent metadata

Field	Value
Publication number	US-12094288-B2
Application number	US-202318491257-A
Country	US
Kind code	B2
Filing date	Oct 20, 2023
Priority date	Jun 21, 2019
Publication date	Sep 17, 2024
Grant date	Sep 17, 2024

Abstract

Disclosed are a system and method for training a neural network associated with a casino table game monitoring system. Synthetic images of objects extracted from a virtual table game environment are used to train a machine learning model, which is deployed to a casino table game monitoring system to monitor one or more physical objects relative to a physical gaming table in a physical table game environment.

First claim

The invention claimed is:

1. A method comprising:
receiving, from a casino table game monitoring system, specification data related to a physical object positioned relative to a physical gaming table;
generating, by a processor based on the specification data, synthetic image data based on a virtual scene that depicts a virtual object relative to a virtual gaming table, wherein the virtual object is modeled in the virtual scene using the specification data, and wherein the virtual gaming table is modeled in the virtual scene according to known dimensions of the physical gaming table;
training, by the processor, a machine learning model using the synthetic image data; and
deploying, by the processor via a communications network, the machine learning model to the casino table game monitoring system to monitor the physical object relative to the physical gaming table.

2. The method of claim 1 , wherein the specification data includes the known dimensions of the physical gaming table.

3. The method of claim 1 , wherein the specification data comprises one or more of images of a casino chip, designs of a casino chip, a texture model of a casino chip, a three-dimensional mesh model representing a casino chip, a color of a casino chip, a casino chip color pattern, a casino branding, a game branding, an image of a bet zone, an electronic design of a table felt, an image of a table felt, and a color of a table felt.

4. The method of claim 1 , wherein the specification data comprises image data indicating one or more of table lighting, color reflections, and shadows associated with a gaming environment in a casino in which the physical gaming table is located.

5. The method of claim 1 , wherein generating the synthetic image data based on the specification data comprises:
converting, by the processor in response to analysis of the specification data, a felt color of the physical gaming table to a white background color, wherein the felt color is indicated by the specification data; and
automatically changing, by the processor, the white background color in each of a plurality of images captured of the virtual scene for the synthetic image data.

6. The method of claim 1 , wherein receiving the specification data comprises receiving a live image feed of the physical gaming table, and wherein generating the synthetic image data comprises:
detecting, by the processor via analysis of the live image feed, the physical object positioned on a game surface of the physical gaming table; and
mapping, by the processor, the detected physical object to a three dimensional mesh modeled from the known dimensions of the physical gaming table.

7. The method of claim 6 , wherein the virtual object is positioned at known coordinates on a game surface of the virtual gaming table, wherein the game surface of the virtual gaming table is modeled from the game surface of the physical gaming table depicted in the live image feed.

8. The method of claim 6 , further comprising superimposing, by the processor via the mapping, an augmented reality overlay over the live image feed.

9. The method of claim 6 , further comprising:
rendering, by the processor according to a view of a virtual camera, the virtual scene;
capturing, by the processor via the virtual camera, image data of the virtual scene;
extracting, by the processor in response to analysis of the image data of the scene, the synthetic image data; and
storing, by the processor, the synthetic image data in a set of ground truth data for the machine learning model.

10. The method of claim 9 , wherein the training comprises training, by the processor, the machine learning model using the synthetic image data from the set of ground truth data.

11. The method of claim 9 , wherein the virtual camera is positioned in the virtual scene to mimic a view of a real camera, positioned relative to the surface of the physical gaming table, to capture the live image feed.

12. The method of claim 11 , wherein the real camera is a depth sensing camera, wherein the specification data includes depth sensing data of the physical gaming table included with the live image feed, and wherein generating, based on the specification data, the synthetic image data of the virtual scene comprises modeling, by the processor based on the depth sensing data, the virtual gaming table in the virtual scene.

13. The method of claim 12 , wherein the virtual camera has a depth of field feature and wherein generating the synthetic image data comprises blurring, using the depth of field feature, an appearance of the virtual scene based on a distance, determined from the depth sensing data, of the real camera to one or more of the physical object or the physical gaming table.

14. A system comprising:
a network communications interface configured to communicate via a communications network; and
a processor configured to execute instructions, which when executed cause the system to perform operations to
receive, from a casino table game monitoring system, specification data related to a physical object positioned relative to a physical gaming table,
generate, based on the specification data, synthetic image data based on a virtual scene that depicts a virtual object relative to a virtual gaming table, wherein the virtual object is modeled in the virtual scene using the specification data, and wherein the virtual gaming table is modeled in the virtual scene according to known dimensions of the physical gaming table,
train a machine learning model using the synthetic image data, and
deploy, via the communications network, the machine learning model to the casino table game monitoring system to monitor the physical object relative to the physical gaming table.

15. One or more non-transitory, computer-readable storage media having instructions stored thereon, which, when executed by a set of one or more processors, cause the set of one or more processors to perform operations comprising:
receiving, from a casino table game monitoring system, specification data related to a physical object positioned relative to a physical gaming table;
generating, by a processor based on the specification data, synthetic image data based on a virtual scene that depicts a virtual object relative to a virtual gaming table, wherein the virtual object is modeled in the virtual scene using the specification data, and wherein the virtual gaming table is modeled in the virtual scene according to known dimensions of the physical gaming table;
training, by the processor, a machine learning model using the synthetic image data; and
deploying, by the processor via a communications network, the machine learning model to the casino table game monitoring system to monitor the physical object relative to the physical gaming table.

Assignees

• Lnw Gaming Inc

Inventors

• Kelly Bryan M
• Eager Terrin
• Lyons Martin S

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/096

  Transfer learning · CPC title

• G06N3/092

  Reinforcement learning · CPC title

• G07F17/3223

  Architectural aspects of a gaming system, e.g. internal configuration, leader-follower, wireless communication · CPC title