individuals gamble each year
annual US gambling revenue
have legal commercial gambling
TensorIoT utilized AWS to build a machine learning computer vision solution that was capable of recognizing, recording, and counting casino chips to determine what bets were being placed at a specific gaming table.
How can you use existing camera systems to recognize, record, and count casino chips to monitor betting taking place at a gaming table?
We used Machine Learning models in AWS SageMaker to identify and process visual images with computer vision.
Casinos have long been searching for ways to better prevent cheating and theft. To date, the best solution has been to use rewards cards to track customers’ movements, but this method is inefficient and can be cumbersome for the customer.
The client, a leading gaming company, was ready to explore the world of IoT, to see if smart cameras could better track customer behavior, and even improve the customers’ experience at the casino. TensorIoT was ready and able to take on this challenge.
Counting casino chips is a novel task for computer vision technology. In order to train the machine learning (ML) models to count chips, there were specific constraints on camera choice and position. The presence of colored lights that casinos often use confused the ML models. Also, if a camera was bumped or changed, its ability to identify chips was diminished. It was also a challenge to design the model for handling chip recognition since gamblers don't always place chips in the same precise location on the gaming table.
Using a simulated casino environment in our office, TensorIoT developed a sequence of machine learning models that relied on each other to pre-process images and correctly count the chip total. Then, with the aid of AWS Sagemaker, the ML models were easily deployed to the edge (casino) environment using AWS Greengrass’s ML resource capability. Proper tabulation of chips now enables accurate bet tracking and monitoring without requiring human intervention.