PhysioNet - Digitization of ECG Images

Overview

This competition challenges you to build models that can extract time-series data from images of Electrocardiograms (ECGs). These ECG images often come from scans or photographs and can have various artifacts making them difficult for current software to process. Your work will help convert billions of historical ECG images into usable data to train better diagnostic models, ultimately improving cardiovascular care.

Requirements

You will need to build models and submit them through Kaggle Notebooks. Your notebook must run within 9 hours on either CPU or GPU, and internet access will be disabled during the run. You can use freely available external data and pre-trained models. The final submission file should be named 'submission.parquet' or 'submission.csv' and contain the 'id' and predicted 'value' for each ECG image.

Prizes

This competition offers substantial cash prizes, with the 1st place winner receiving $10,000, 2nd place $9,000, and 3rd place $8,000. The prizes continue down to 7th place, with $5,000 awarded to both 6th and 7th place finishers, totaling $50,000 in cash prizes. Participants also earn awards points and medals.