Entity Detection, or ED, is our machine-learning visual recognition software, currently trained to identify, geo-locate, and automatically create a database New Zealand road signs.
While our machine learning algorithm can be applied to almost any other video footage, we trained Entity Detection using a standard ZED 2i stereo camera from StereoLabs.
This enables road signs to be inspected and catalogued for maintenance at highway speeds, reducing inspection team cost by over 70%.
Entity Detection can feed data into any system. When fed into the Nextspace platform, the roadside database can be integrated into a single highway conditions model, including GIS, contours, moisture levels, roadworks, underground utilities, powerlines, rail tracks, accident reports, and so on.
Entity Detection is priced as a combination of initial machine learning–training and a per–mile analysis rate.
Training cost depends on a number of factors:
Complexity of asset recognition. For example, New Zealand roadsigns are designed to stand out from all backgrounds.
Availability of training imagery library. In the case of New Zealand roadsigns this included graphic design files as well as photos and video of actual roadsigns from various angles and in various environmental conditions.
Ease of in-field capture for training. In the case of New Zealand roadsigns, our client had crew constantly driving the roads as part of normal operations.
Team for initial machine learning–training (internal or external). We can supply training resource or teach your team how to train Entity Detection.