Machine learning models can run on the data collected by the sensors. Using our modular software, models can be deployed easily, regardless of which frameworks they have been trained with, e.g. TensorFlow or Pytorch. We automatically optimize them to maximize performance on the embedded hardware.
Our goal is to develop partnerships with cutting edge solutions proposed by key AI players, in order to offer end-to-end technologies to our clients, and we ensure compatibility with their products. Machine learning models can be easily created through a specialized data science platform and then automatically and effortlessly deployed to a fleet of robots.
Our tracking algorithms significantly improve the inference precision of any object detection model. They reduce false positive detections that are common when a drone is flying at high altitudes and therefore, every identifiable object looks small.
We leverage the acceleration capabilities of embedded solutions to improve the video quality, applying enhancements such as real-time software stabilization and noise reduction. We provide a smooth experience even when robots navigate into difficult environments such as in adverse weather conditions.