Computer Vision News - June 2019

I guess it’s not the same type of constraints of real time and FpS. You can use different models than ones that run in real-time that you could use for car applications. Again, there are some differences there. We usually use a mix of a deep learning type of approach also with a traditional computer vision type of approach. This is slightly less complex, as we primarily need to distinguish what's not water, but I guess that’s pretty much the same for self-driving cars .” Shone develops proprietary detection algorithms based on data gathered from the ship sensors AIS, a communications system which people are trying to implement in the autonomous space. They are also working on an augmented reality type of application to empower the crew to navigate more safely in bad weather conditions, which will help during situations in which the crew cannot see an obstacle with the human eye. The technology analyzes an augmented reality object on top of the camera feed projected from a radar target, AIS target, or an element of the map such as buoys or the ship’s pathway. With this technology, the crew gains a better sense of their surroundings. Currently, Shone has nine employees based in San Francisco and three in Europe. For over a year, several CMA CGM ships have been equipped with the technology predicting terabytes of data from those vessels . The system continues to improve with each day. They have already started to deploy to more vessels with the aim toward autonomy. Along the way, they need to develop a lot of technology, some of which they will package and sell as a product. Pauline Luc 21 Computer Vision News Application Shone