Skip to content

ASL Recognition

American Sign Language alphabet recognition in real time, with YOLOv11 trained on a custom dataset.

Category
AI
Confusion matrix from the ASL model evaluation: a heatmap of predicted against true labels for the letters A to Z plus background, with correct predictions concentrated along the diagonal and per-class counts reaching 26.

Overview

This project recognizes the letters of the American Sign Language alphabet from a live webcam feed. The work covered building and annotating a custom dataset in Roboflow, training a YOLOv11 model on it, and running detection on each video frame with OpenCV. The model was evaluated with a confusion matrix across the letters A to Z. It is meant as a learning aid for people studying ASL.

What it does

  • Real-time ASL alphabet recognition using YOLOv11 architecture
  • Custom dataset creation and annotation with Roboflow
  • High accuracy detection with optimized computer vision algorithms
  • Live webcam integration for instant gesture recognition
  • Educational tool for learning American Sign Language

Built with, 6

  • Python
  • YOLO
  • NumPy
  • Scikit-learn
  • OpenCV
  • Roboflow