Braille AI: a convolutional neural network-based mobile application for bidirectional Braille translation in Spanish.

Area of Science:

• Assistive Technology
• Artificial Intelligence in Education
• Human-Computer Interaction

Background:

• Students with visual impairments encounter educational challenges due to a scarcity of Braille instructors and learning resources.
• Bridging communication gaps is crucial for inclusive education.

Purpose of the Study:

• To introduce Braille AI, an Android application for bidirectional Spanish text and Braille translation.
• To provide visual guides for Perkins Brailler operation, assisting educators with limited Braille expertise.

Main Methods:

• Integration of YOLOv8 for Braille symbol detection and a TensorFlow Lite CNN for character classification.
• Training AI models on public datasets (Angelina, DSBI, Kaggle Braille).
• Development of a user-friendly Android interface.

Main Results:

• The symbol recognition model achieved ~97% accuracy.
• Translation processing time was under five seconds.
• User evaluations indicated improved accessibility and reduced preparation time for educators.

Conclusions:

• Braille AI serves as a portable, cost-effective solution for communication between educators and visually impaired students.
• The application highlights AI's potential to advance inclusive education, particularly in Spanish-speaking regions.