Machine Learning Surf Cameras
The surf zone on a popular beach is a particularly dynamic environment with waves, wind and currents interacting with surfers and bathers. Monitoring this is tricky, whether you’re a lifeguard charged with keeping people safe, a surfer looking to best time a session or anyone involved in coastal development and long term planning.
At Surfline Labs we’ve built a robust machine learning monitoring system for this environment, using a bespoke deep learning object detection network and a coupled tracking system, we can monitor all water users and waves in real-time. Being able to track every wave surfed, and where they’re occurring, is useful but many of the questions that arise aren’t immediately answerable. How long is the average wave? How far do surfers sit from the shore? Where exactly are the waves breaking most consistently?
All of these questions need answering in real-world space on the map, relative to the beach you’re actually surfing, not pixel space on a camera frame. To respond, we need a robust method to convert between the two and that’s what we’ve been adding to our system.
With our new system we can take live camera feeds from any beach, detect all relevant activity and convert to map coordinates. Constructing something like the heat-map above is then relatively trivial.
We’ll be launching a live demo soon.