It is provided a method of detecting human drowning, comprising: attempting to detect humans in a sequence of underwater images taken by a single camera, for identifying humans-in-water candidates in the images, said detection using at least a machine learning algorithm, tracking humans-in-water candidates throughout this sequence, and detecting human drowning risk. It is also provided a system for detecting human drowning, comprising at least one underwater camera configured to take at least a sequence of underwater images, the system being configured to attempt to detect humans in the sequence of underwater images, for identifying humans-in-water candidates in the images, said detection using at least a machine learning algorithm, track humans-in-water candidates throughout this sequence, and detect human drowning risk.