AI drowning detection to save more than 7,000 lives per year

More than 20 people die every day from drowning in swimming pools*, and most of them are kids
This drowning detection AI is combining machine learning and computer vision to detect drowning people in real time Another meaningful use of #AI, this time developed by Rodman Louis
Detail of the technologies used:
✓ Background Subtraction (Opencv)
✓ Single Pose Estimation (Resnet)
✓ Kalman Filter Tracking (SORT)
✓ Human Activity Recognition (LSTM) (~5 fps on GTX 1080 , Written in Python & Pytorch)
Similar code available in GitHub:
SwiMMinD ( is also working on a similar AI solution but to provide automatic and meaningful feedback to swimmers at any level. SwiMMinD also focuses not only on underwader detection but above the water where the fluidity of the water is unpredictable. The goal is for anyone to record his swim via his device (mobile phone etc) and to receive automatic stroke corrections.
* according to National Safety Council reports, statistics for USA only

Leave a Comment