Vital Extraction Challenge
1st Postion (Gold Medal)
Bagged a Gold Medal among 20+ participating IITs
- Employed a YOLOv8 object detection model to extract essential parameters from ECG monitor images
- Innovatively devised a classification-based segmentation approach to detect the screen corner of ECG monitors, enhancing accuracy in subsequent data extraction
- Conducted in-depth exploration of OCR techniques, including parseq, ABINet, and PaddleOCR along with pioneering a novel technique for generating Heart rate and SpO2 graphs from ECG image data