In the last decade deep learning techniques have replaced classical computer vision algorithms in many tasks. Insurance is one of the many successful applications of deep learning in image recognition.
Flair-tech is supporting Exclusive Garage in designing a fully automatic car damage recognition system based on deep learning. The predictive model is trained to accurately detect and localize scratches and dents on a car’s body.
The electric power consumption of a railway network is often available as an aggregate measurement, with no insight into the contribution of individual trains. On the other hand, on-board measurement devices available on the market do not provide a cost-efficient and scalable solution for large networks.
Flair-tech has supported NRG.it in developing a scalable, low-cost measurement system based on machine learning and GPS. The underlying predictive model is able to infer the consumption of individual trains of a railway network given their instant GPS coordinates and information on their paths.
When people and objects have to be accurately located within indoor, narrow or irregularly shaped environments, GPS tracking is not a reliable option. On the other hand, sophisticated antenna-based locators are often costly black boxes, requiring heavy customization to be effective.
Flair-tech has supported Dotdotdot to design and develop a low-cost, easily customizable indoor positioning system, based on machine learning and Bluetooth signal. We have trained the underlying model to map the signal measured by a chain of receiver nodes to the location of a person/object within a complex environment.
VERIMA is a software solution by Witapp that allows a surgical team to plan their surgery carefully. Via a mixed reality visor, CT body scans can be viewed and interacted with as 3D holograms.
The deep learning core of VERIMA, that Flair-tech has proudly engineered and trained, is able to process raw CT scans and automatically highlight bones, vessels and organs, exploiting state-of-the-art semantic segmentation techniques.