Khoi Tran Dinh Anh

Khoi Tran Dinh Anh

AI Learning

Vietnamese Card Games

Developed a computer vision-based system to recognize card sequences and individual cards in popular Vietnamese card games such as Tiến Lên, Ba Cào, and Xì Dách. The system combines deep learning techniques and custom image processing algorithms to accurately detect and classify playing cards, identify card sequences, and assess player hands.

  • Card Detection and Classification: Trained a custom model using YOLO (You Only Look Once) for real-time detection of individual cards, card ranks, and suits from game environments with various card layouts and orientations.

  • Sequence Recognition: Implemented custom algorithms to detect and validate card sequences (e.g., straights, three-of-a-kind, pairs) based on game rules for Tiến Lên and other card games. The system was designed to handle complex combinations and multiple cards within a player’s hand.

  • Optimized for Real-Time Performance: Ensured fast and accurate recognition through the integration of optimized image preprocessing techniques, allowing the system to be used in real-time applications such as digital card tables or automated game systems.

  • Game-Specific Customization: Tailored the detection system to the unique rules and card combinations specific to each game (e.g., handling the special card sets in Ba Cào and counting strategies in Xì Dách).

Technologies Used: Python, OpenCV, YOLOv5, PyTorch, custom rule-based algorithms for game-specific logic.

This project demonstrates expertise in computer vision, deep learning, and the application of custom algorithms to solve complex pattern recognition problems in real-world gaming environments.