TECHNOLOGY

Enabling the next generation of robots and autonomous vehicles

Univrses’ technologies are based on four years of intense work done by our team of scientists and engineers. We work in a variety of industries, and have particularly high expertise within 3D computer vision and 3D AI. We are experts in providing solutions that enables accurate analysis of the surrounding environment by the use mobile cameras. Today, our computer vision and machine learning technologies are amongst the most advanced in the world.

LOCALIZATION & MAPPING

WHAT IS LOCALIZATION AND MAPPING?

By adding cameras onto equipment, you can track the equipments’ movements while at the same time mapping its surroundings. Mapping a given environment creates a virtual 3D representation that allows equipment to localize and communicate 3D poses in relation to other equipment. This technology allows you to map both very small spaces and also large indoor or outdoor environments. It can be used in many practical situations, for instance by putting cameras on a vehicle to help it map the world and localize in it. Mapping technology can be applied on numerous objects, but every situation requires different approaches to the specific problem.

HOW DOES UNIVRSES USE LOCALIZATION AND MAPPING?

At Univrses, we have a strong, experienced team and mature technology in and around localization and mapping. Our system can work with several moving agents in the same local space. It has high performance, high accuracy and high robustness. The solution can be tweaked to a lot of different scenarios and a lot of different sensors. There are several PhDs and MScs working in this field at Univrses. Our expertise is applicable in several industries. We are working with autonomous driving, as well as home robotics, industrial robotics, 3D reconstruction for warehousing solutions and more.

TECHNOLOGIES
Visual odometry
Visual inertial odometry
Multi-camera odometry
SLAM
VI-SLAM
Lidar-SLAM
Multiagent Mapping
Cloud Mapping
3D Reconstruction
Monocular Vision
Stereo Vision
RGBD Vision
​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​

SPATIAL DEEP LEARNING

WHAT IS SPATIAL DEEP LEARNING?

Deep learning enhances localization and mapping technology with a semantic interpretation of the environment. By translating mapping sensor data to human interpretable information (e.g. by localizing specific objects) it enables autonomous platforms to operate in co-shared space with humans. Deep learning technology plays a very important role in the development of autonomous driving, since it makes it possible for a vehicle to spot meaningful landmarks..

HOW DOES UNIVRSES WORK WITH SPATIAL DEEP LEARNING?

Univrses applies deep learning in 3D. Our entire team is well versed and experienced in working with 3D geometrical optimization problems in computer vision and deep learning. By applying 3D deep learning, we can really make sense of the surroundings. For instance, when we work with automotive, this technology can build semantic maps of the environment. The same goes for understanding where there is a door, a house, a road sign etc. Univrses has a team of computer vision and machine learning scientists with an extensive history of publications in the field, working with spatial deep learning and computer vision technologies.

TECHNOLOGIES
Semantic Mapping
Mono-Depth
Semantic Segmentation
Object Detection
Object Tracking
Domain Adaptation
​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​

SENSOR FUSION

WHAT IS SENSOR FUSION?

Sensor fusion is using sensory data from numerous sources, combined into one mutual result. Making sense of this data is an integral part of solving robust 3D perception. In some cases, redundancy is needed, and an important part of the sensor fusion is making sure that several sensors agree on the same conclusion. In other cases, it is about perfecting an optimization by looking at the output from a multitude of sensors.

HOW DOES UNIVRSES WORK WITH SENSOR FUSION?

Univrses has extensive experience from sensor fusion using a lot of different types of sensors. Depending on the specific problem, we work either with tightly integrated solutions, where several sensory data is optimized in the same optimization problem, or with different types of filters when late sensor fusion is more suitable for the problem space. We have experience from working with many different sensors, but even if your specific sensor would be new to us, our team of experts will find the right state-of-the-art solution to help you anyway. The Univrses technologies and team are well suited to work on sensor fusion problems.