YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Currency Detection Model

A lightweight currency detection model trained using the Ultralytics YOLO architecture for assistive technology applications such as EyeCure, a system designed to help visually impaired users identify Indian currency notes using a camera.


Model Details

  • Architecture: YOLO26n
  • Framework: Ultralytics YOLO
  • Task: Object Detection
  • Classes:
Class
₹10
₹20
₹50
₹100
₹200
₹500

The model is optimized for real-time inference and lightweight deployment on edge devices.


Training Configuration

  • Epochs: 100

  • Image Size: 640

  • Batch Size: 16

  • Dataset Split:

    • Train: 70%
    • Validation: 20%
    • Test: 10%

Training was performed on a GPU (Tesla T4).


Evaluation Metrics

Overall validation performance:

Metric Value
Precision 0.977
Recall 0.926
mAP@50 0.966
mAP@50-95 0.868

Per-Class Performance

Currency Precision Recall mAP@50 mAP@50-95
₹10 0.990 0.915 0.982 0.935
₹20 0.972 0.957 0.980 0.889
₹50 0.990 0.977 0.993 0.898
₹100 0.959 0.930 0.961 0.800
₹200 0.990 0.933 0.978 0.901
₹500 0.959 0.845 0.903 0.784

Inference Speed

Stage Time
Preprocess 0.3 ms
Inference 2.8 ms
Postprocess 0.6 ms

This enables real-time detection in live camera applications.


Example Use

from ultralytics import YOLO

model = YOLO("currency_model.pt")

results = model("currency_image.jpg")
results.show()

Use Case

This model was developed as part of the EyeCure assistive system, which provides:

  • Real-time currency recognition
  • Voice feedback for visually impaired users
  • Offline edge deployment

Limitations

  • Performance may drop in extremely low lighting conditions.
  • Detection accuracy may decrease for heavily folded or partially visible notes.

Future Improvements

  • Improve multi-currency detection in a single frame
  • Expand dataset with real-world scenarios

Author

Abdul Saboor AIML Engineering Student EyeCure Project

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support