Automated rip current detection with region based convolutional neural networks

This paper presents a machine learning approach for the automatic identification of rip currents with breaking waves. Rip are dangerous fast moving water that result in many deaths by sweeping people out to sea. Most do not know how recognize order avoid them. Furthermore, efforts forecast hindered lack observations help train and validate hazard models. The presence web cams smart phones have made video still imagery coast ubiquitous provide potential source current observations. These same devices could aid public awareness currents. What is lacking method detect or absence from coastal imagery. provides expert labeled training test data We use Faster R–CNN custom temporal aggregation stage make detections images videos higher measured accuracy than both humans other methods detection previously reported literature. • Evidence region based object detectors applicable amorphous ephemeral objects such as Analysis showing above existing published methods. Data sets testing.