ryansakurai/sign-language-recognition

Brazilian Sign Language alphabet recognition project made for an Artificial Intelligence class.

Brazilian Sign Language Alphabet Recognition - Artificial Intelligence Project

Concept

A project aimed at recognizing Brazilian Sign Language alphabet signs through gestures captured by the webcam.

How to Use

• After downloading all the files, install the dependencies with "pip install -r requirements.txt". After that, simply enter this command to start the program directly in the terminal: "python main.py".
• Make hand signs and the program will display the detected letters in real-time on the screen.

Resources Used

The model was created using Google's Teachable Machine and the program was developed using Python and the following libraries:

• OpenCV - video capture and manipulation;
• MediaPipe - hand processing;
• Keras - AI model;
• NumPy - numerical representation of data.

Known Issues

• Due to the use of static images to recognize gestures, it was necessary to use static representation for signs that include movements.
• The program works more effectively with a smooth background.
• Some signs have significantly lower success rate than others, as shown in the following graph:

Authors

• Guilherme Bortoletto
• Ryan Sakurai
• Vinicius Castro
• Marcelo Pirro
• Caike dos Santos

(Computer Science students at UFSCAR - Federal University of São Carlos)