Lausanne, Switzerland, September 07, 2021 – Picterra – the Swiss-based geospatial intelligence platform – today announced the launch of multi-class detectors, giving users a more powerful way of applying machine learning to geospatial images. This is the latest innovation in a suite of enterprise upgrades to the platform including user management and permission control.

Previously, Picterra users created one model per object class they wished to identify in an image. For example, one for pavement cracks and another for potholes. With the availability of multi-class detectors, users only have to train one model to detect multiple object classes. Early performance tests indicate that the use of multi-class detectors increases the per-model accuracy by up to 25% (depending on the use case) and triples the annotation speed. 

“This is a game-changer for Picterra users,” says CTO & Co-Founder Frank de Morsier. “Our goal is to make it easy for clients to get insights from geospatial imagery in a fast way. This innovation delivers on that promise. I am proud of our team for continuing to deliver powerful deep learning features that anyone can use, regardless of skill level.”

Multi-class detectors are available on all Picterra plans. Get in touch with Sales to learn more.