AI / MLAytasense TechnologiesAug 2024 to Dec 2024
Real-time iris-to-pupil ratio detection
The iris-to-pupil ratio is a possible non-invasive marker for neurological conditions. We built a pipeline that measures it from a camera frame in real time.
My roleR&D intern, team of four
Precision 0.989 and recall 0.949 on validation; paper submitted to Biomedical Materials and Devices.
- 0.987
- mAP@0.5
- 0.989
- precision
- 0.949
- recall
The problem
Changes in the iris-to-pupil ratio may signal neurological conditions such as Alzheimer’s and Parkinson’s. Measuring it should be fast, cheap and non-invasive.
What we built
A pipeline from camera frame to ratio: OpenCV Haar cascades locate the eyes, CLAHE and reflection removal clean each crop, and a fine-tuned YOLOv5s model detects the iris and pupil. A Streamlit app runs it from a live camera or an upload.
Results
After 100 epochs: mAP@0.5 0.987, precision 0.989, recall 0.949, mAP@0.5:0.95 0.903 on the validation set.
Context
R&D internship at Aytasense Technologies (August to December 2024), done as a team of four students from RVITM.
- Python
- YOLOv5
- OpenCV
- Streamlit