Extreme Video Analytics Experiment: Detecting Kitesurfers

Source: Deep Learning on Medium

Extreme Video Analytics Experiment: Detecting Kitesurfers

Now deep learning software algorithms are capable of recognizing almost any object or event that you need on any background. You just define the task and software starts collecting information for you from available videostreams. And the limit today is not capabilities of the software, but your imagination. To prove this, we decided to give this extreme case: detecting amazing kitesurfers and their kites.

Seascape is traditionally one of the most difficult cases for video recognition software. Using multi-purpose recognition platform BitRefine Heads we try to detect and count kite canopies, track their movement and save their colors to be able to do search using various properties. After that we try to detect kitesurfers themselves and track them moving across the water.


Although scene is pretty dynamic, we can detect and mark kite canopies.