Braille-lens is a mobile application developed to recognize and translate Filipino (Tagalog) Braille characters and contractions into text with Text-to-Speech.
Developed as a thesis under the Department of Computer Studies, Cavite State University – CCAT Campus.
Braille-Lens: Tagalog Braille Alphabet and Contractions Recognition with Text-to-Speech Using YOLOv8
Researchers: Jyra Mae Celajes, Crestalyn Luardo, Louie Jenn Jaspe
Course: Bachelor of Science in Computer Science
-
🇵🇭 Filipino Braille Recognition
Detects Grade 1 and Grade 2 Braille characters using a YOLOv8-based object detection model. -
📷 Capture / Import / Sample
Capture Braille images using the camera, import from gallery, or test with sample images. -
🔄 Grade Options
Toggle between Grade 1, Grade 2, or both models for Braille detection. -
🧠 Supported Braille Patterns
- English Grade 1 alphabet
- Filipino alphabet (includes ñ and ng)
- Capital sign, number sign
- Digits (1–9)
- Filipino-Tagalog One-cell contractions (alphabet and non-alphabet)
- Filipino-Tagalog One-cell part words
- Filipino-Tagalog Two-cell contractions (alphabet and non-alphabet)
-
🟤 Real and Simulated Braille Support
Works with both:- Simulated Braille (black dot representations)
- Written/embedded Braille (captured with lighting that shows dot shadows) See samples here: Sample Input Images
-
📝 Braille to Text + TTS
Converts detected Braille into Tagalog text and reads it aloud using Text-to-Speech. -
📖 Dictionary
Built-in searchable reference for Filipino Grade 1 and Grade 2 Braille symbols. -
🎬 User Guide
“How to Use Braille-lens” section includes a demo video and instructions. -
⚙️ Settings
- Adjust TTS pitch and rate
- Preview speech output
- Switch between light and dark theme
-
📊 Result Screen
Displays detection results, confidence score, and an information sheet. -
ℹ️ About Page
Includes background on Braille, Filipino Braille rules, Braille grades, app purpose, and resources.
Watch the demo of Braille-lens in action:
| Model Type | Dataset | Precision | Recall | mAP50 | mAP50-95 |
|---|---|---|---|---|---|
| Grade 1 Braille | Custom Filipino Braille Dataset | 0.97 | 0.97 | 0.99 | 0.84 |
| Grade 2 Braille | Custom Filipino Braille Dataset | 0.98 | 0.98 | 0.99 | 0.80 |
Evaluated on over 21,000 annotated Braille cell images with 247,050 annotations, covering 67 classes for Grade 1 and 89 classes for Grade 2. Trained using YOLOv8-small (YOLOv8s) and deployed via TFLite for Android inference.
- 🔖 Latest Release: v1.0
- 📁 APK:
braille-lens-v1.0.apk - 📱 Platform: Android
- ⚙️ Built with: Kotlin, Jetpack Compose, YOLOv8, PyTorch, TFLite
University: Cavite State University – CCAT Campus
Department: Department of Computer Studies
Thesis Members:
- Jyra Mae Celajes
- Crestalyn Luardo
- Louie Jenn Jaspe
- 🧠 YOLOv8 object detection (PyTorch)
- 📱 Kotlin + Jetpack Compose
- 🗣️ Android Text-to-Speech
- 🗂️ TFLite integration
- 🖼️ Roboflow, Canva, and Braille dataset curation
This project is for academic and educational use only.
Please contact the authors for permission regarding reuse or publication.
- Ultralytics YOLOv8
- Android Jetpack Compose & TTS APIs
- Instruction Manual for Filipino Braille Transcription
- Roboflow and Canva
- Thesis panel, advisors, and all testers
Found a bug or have suggestions?
Please submit an issue here → GitHub Issues