VeeringAnalytics

Over the past few years several cost-effective instrument systems have come to market that are capable of collecting meaningful data quickly and easily from small keel boats and dinghies, as well as to measure loads. To date there has been little work on capturing the position of controls to corelate with these loads and performance metrics.

Veering Analytics’ Visual String Pot algorithm works by simply tracking two distinct markers, one at a fixed location and of a known length, and the other attached to the point of interest. These markers are tracked in a series of photographs using commercially available cameras (such as a GoPro).  The algorithm provides a timeseries output at the frequency the photos where taken.

As with many computer visions tasks, Visual String Pot is a computationally heavy task. By leveraging Facebooks Fast Accurate Similarity Search (FAISS) algorithms Visual String Pot is able to run on (relatively) available hardware and requires minimum use of expensive and difficult to obtain GPU compute time. In some situations, up to 10,000 images can be processed in under half an hour on a standard 12th gen core i9 machine. Veering Analytics has the computer power available onsite to meet the requirements of all our proprietary programs. Simply upload the days images and let us return a simple CSV time series or have it integrated with your boat’s existing log files or analysis.

A gopro and power bank deployed to track mast gate position on an etchells

Raw images prior to processing

Sample images showing the tracked dots and the calculated distances.