A smart, AI-powered wearable designed to assist visually impaired individuals in identifying Indian currency denominations and detecting damaged or fake notes — all processed offline on a Raspberry Pi 5 using YOLOv8.
THEIA is a smart wearable device designed to help visually impaired individuals identify Indian currency and assess note condition in real time. Built with a Raspberry Pi 5, Camera Module 3, and YOLOv8 AI models, it provides offline recognition, ensuring accessibility, independence, and fraud prevention.
This device is enclosed in a 3D-printed body with tactile switches for performing different actions, including currency and damage detection. It delivers results through earphones or speakers, offering a completely hands-free experience.
🧠 Detection and feedback occur in under 2 seconds, entirely offline.
- 💰 ₹35,000 Institutional Grant from Rajagiri School of Engineering & Technology
- 🇬🇧 UK Design Patent for wearable design innovation
- 🎓 Funded and recognized as a college research project in assistive technology
| Feature | Description |
|---|---|
| 💵 Real-Time Detection | Identifies Indian currency denominations in <2 seconds using YOLOv8. |
| 🩸 Damage & Screen Detection | Detects torn, folded, or digital note images to prevent misuse or fraud. |
| 🖲️ Hands-Free Control | Tactile switches on the 3D-printed shell let users toggle between modes easily. |
| 🗣️ Audio Feedback | Outputs denomination and condition via earphones or speakers using Pico2Wave TTS. |
| 🌐 Offline Operation | Works entirely without internet using Raspberry Pi 5 and on-device AI. |
| Component | Technology |
|---|---|
| Programming Language | Python |
| Model | YOLOv8 (Ultralytics) |
| Computer Vision | OpenCV |
| Hardware | Raspberry Pi 5 (4GB RAM), Raspberry Pi Camera Module 3 |
| Enclosure | Custom 3D-Printed Body with Buttons |
| Power | 2000mAh Rechargeable Battery |
| Audio Output | Pico2Wave (TTS), Speaker / Earphones |
| Dataset Tool | Roboflow |
-
Currency Detection Dataset → Roboflow Link
-
Damage Detection Dataset → Roboflow Link
-
Total Images: 10,000+
-
Classes:
10_new, 10_new_folded, 10_old, 10_old_folded, 100_new, 100_new_folded, 100_old, 100_old_folded, 20_new, 20_new_folded, 20_old, 20_old_folded, 200_new, 200_new_folded, 50_new, 50_new_folded, 50_old, 50_old_folded, 500_new, 500_folded, non-currency, screen_image, 10_new_damaged, 10_old_damaged, 100_new_damaged, 100_old_damaged, 20_new_damaged, 20_old_damaged, 50_new_damaged, 50_old_damaged
Camera (Input)
↓
OpenCV (Preprocessing)
↓
YOLOv8 (Currency & Damage Detection)
↓
Raspberry Pi 5 (Processing)
↓
Pico2Wave (Voice Output)
↓
Earphones / Speaker (User Feedback)
🧩 Hardware Integration:
- Raspberry Pi Camera captures input frames.
- OpenCV handles preprocessing.
- YOLOv8 performs both currency classification and damage detection.
- Pico2Wave generates real-time voice feedback.
- The 3D-printed shell houses buttons for action control and a 2000mAh battery for portability.
- User holds a note in front of the camera.
- YOLOv8 model detects the note and identifies denomination and condition.
- Audio feedback announces the denomination and whether the note is damaged or fake.
- Switch buttons allow toggling between different detection modes.
- Support for multiple currency types.
- Integration of voice-based commands.
- Cloud synchronization for dataset updates and analytics.
- Companion mobile app for configuration and updates.
| Name | |
|---|---|
| Abel George Stanley | u2204003@rajagiri.edu.in |
| Anandhakrishnan J | u2204013@rajagiri.edu.in |
| Anjith Saju | u2204016@rajagiri.edu.in |
| Namit Rajeev | u2204045@rajagiri.edu.in |
| Swathi S | u2204064@rajagiri.edu.in |
Institution: Rajagiri School of Engineering & Technology Guided by: Mr. Mathews Abraham
- 🌐 Website: https://theia-pied.vercel.app/