AI Solution for Analyzing and Streaming Casino Card Games in Real-Time
In the world of casinos, strict adherence to security protocols and constant game analysis are paramount. Effective game analysis provides insights into table load, potential fraud schemes, and employee behavior during unusual circumstances. This awareness is critical for achieving business success.
To enhance security and minimize human error, a significant amount of analysis work can be automated. Recent advancements in Computer Vision and Machine Learning surpass human capabilities in speed and attention to detail, making automation an attractive solution
Project Challenge
The client approached Requestum with an intriguing idea to develop software for real-time analysis and streaming of casino card games. Our mission was to design a custom AI solution that could autonomously track casino players, playing cards, chips, and game actions. The project presented several specific challenges, such as:
Moreover, the onset of COVID-19 restrictions introduced an additional challenge in ensuring the effective functioning of the player recognition algorithm with face coverings.
Real-time game analysis
Ensuring immediate processing and analysis of game activities.
Security
Maintaining a secure environment for data transmission and analysis.
Precision
Achieving high accuracy in detecting players and actions.
Action recognition
Identifying various game actions accurately.
Player recognition
Identifying various game actions accurately.
Synthetic dataset utilization
Creating a synthetic dataset due to the lack of real-life datasets.
-
Real-time game analysis
Ensuring immediate processing and analysis of game activities.
-
Security
Maintaining a secure environment for data transmission and analysis.
-
Precision
Achieving high accuracy in detecting players and actions.
-
Action recognition
Identifying various game actions accurately.
-
Player recognition
Identifying various game actions accurately.
-
Synthetic dataset utilization
Creating a synthetic dataset due to the lack of real-life datasets.
Moreover, the onset of COVID-19 restrictions introduced an additional challenge in ensuring the effective functioning of the player recognition algorithm with face coverings.
Solution
To address these multifaceted challenges, we integrated a custom AI algorithm into a peer-to-peer WebRTC streaming solution, facilitating secure and efficient data transfer to a dedicated web dashboard for casino management. The custom AI solution employs computer vision algorithms to:
Detect players and dealers
Assigning unique identifiers (IDs) to each participant.
Recognize game and table equipment
Identifying playing cards, playing chips, and the number of cards in players’ hands.
Monitor game actions
Tracking the start and end of rounds, along with player decisions like insurance, doubling down, and surrender.
Real-time video enhancement
Editing video feeds to include intuitive visualizations during live streaming.
-
Detect players and dealers
Assigning unique identifiers (IDs) to each participant.
-
Recognize game and table equipment
Identifying playing cards, playing chips, and the number of cards in players’ hands.
-
Monitor game actions
Tracking the start and end of rounds, along with player decisions like insurance, doubling down, and surrender.
-
Real-time video enhancement
Editing video feeds to include intuitive visualizations during live streaming.
Core Features
Our solution includes several key features that enhance its functionality:
-
Seamless real-time streaming via WebRTC
The AI solution enables seamless and secure peer-to-peer streaming of video data directly from the gaming room camera to authorized users, ensuring instant access to live gameplay.
-
Player & dealer detection
Each player is assigned a unique identifier through advanced face tracking and recognition technology. The dealer is also detected and assigned a specific role. Our algorithm effectively distinguishes individuals, even when their faces are obscured by medical masks.
-
Video editing
The solution enhances the original video feed by incorporating graphic elements in real time, highlighting critical player and dealer information, as well as the values of chips and cards. The result is a visually informative video stream that aids in immediate decision-making.
-
Table tracking
The system meticulously monitors the playing table, accurately assessing the quantity and value of chips and playing cards held by each player, providing essential insights into game dynamics.
-
Action recognition
The system precisely captures the start and end of each game round, along with key player actions such as insurance, doubling down, and surrender. All relevant data is thoroughly documented for comprehensive analysis.
Tech Stack
Python
TensorFlow
YOLO
OpenCV
RabbitMQ
Postgres
Blender 3D
Results
As a result, we implemented a solution that proved to be an invaluable asset for our client. It provides casino owners with critical insights into table load, employee behavior, and potential security vulnerabilities. The software not only enhances security measures but also equips our client with a competitive edge, facilitating more efficient casino operations through improved analysis of games, players, and staff.
