What we can deliver
Hardware Edge Platforms
Easy to integrate low power edge based platforms optimised for AI vision based solutions.
Neural Network Development
Develop your Neural Network from data to training to inference.
Full Edge AI solution
Connected Hardware Edge Platform running your developed Neural Network.
EdgeVision™ visual IoT solution for real-time processing of visual data generated at the network edge
Ubotica’s EdgeVision™ AI-powered visual IoT platform enables the direct real-time processing of visual data generated at the network edge. This truly embedded solution with integrated image sensor consumes visual data to directly produce actionable output in real time, thereby reducing bandwidth and power requirements.
Ubotica’s custom software toolchain focuses on application lifecycle, with unique technology that enables the efficient deployment of Neural Network updates to edge devices over extremely low bandwidth connections. This directly addresses the important requirement of being able to improve the performance of field-deployed NN-based IoT solutions.
Truly embedded solution – no need for PC tether
Low power envelope and simple SPI control protocol
Powered by Intel Movidius Myriad X, with industry-leading inferences/second/Watt performance
Full integration with Intel’s free OpenVINO AI toolchain for model conversion and optimisation
Enabling Artificial Intelligence in-orbit for image processing and neural network inference applications
Ubotica’s range of hardware and software solutions addresses the New Space market by enabling the application of AI in-orbit using state-of-the-art hardware and easy to use custom development tools. Our technologies span the gamut of platforms from nanosats and cubesats up to large Earth Observation platforms, enabling image processing and neural network inference applications within thermal and power budgets as low as 1W.
Our patent pending technology, custom developed for New Space applications, facilitates the efficient update of neural networks over bandwidth limited uplink channels. In combination with the in-flight software reconfigurability of the hardware solution, this directly opens the path to satellite-as-a-service applications, maximising platform utilisation by enabling application switching and mission re-purposing.
Hardware platform coupled with easy-to-use software framework
Software-reconfigurable hardware image processing pipelines and graphical pipeline development tools
Radiation-characterised COTS solutions
Video Orchestrated Freight Management system succeeding RFID
Lenovo and Ubotica collaborated to design, architect and implement the Video Orchestrated Freight Management system that monitors and optimises stock movement and placement within Lenovo’s warehouse environments.
Operating on Lenovo’s Edge Server technology, and using Intel’s Artificial Intelligence OpenVINO framework to accelerate inference, the system processes video streams in parallel from multiple cameras to disposition the content of forklifts automatically and efficiently. To achieve this, we implement a concept of ‘Virtual Gates’ and use Data Matrix codes in the warehouse.
Complementary technology running on Windows based laptops on board the forklifts automates the “Put Away” and “Pick Up” processes.
Overall the system interfaces to backend ERP systems.
Video Orchestrated Freight Management is operational at facilities in the US and China.
Images digitally assessed with improved accuracy, increased speed, and at a lower cost
Diabetic Retinopathy (DR) is the leading cause of vision loss in adults. National screening programs are in place in Ireland (Diabetic RetinaScreen) and other geographies. This project aims to use Artificial Intelligence (AI) to digitally (computer vision) assess screened retina images for improved accuracy (reduce error rate), increased speed, and at a lower cost.
The solution, which was developed in conjunction with Intel, and with German research organisation fortiss, runs on a standard Windows 10 PC, uses Intel’s OpenVINO framework for AI and Intel’s USB Neural Compute Stick for AI acceleration.
The project achieved improved accuracy of retina image assessment (computer vision versus human vision) in the process to determine if Diabetic Retinopathy symptoms exist or not. Reduce the numbers of inaccurate (error) image readings.
Ultimately the computer vision capability can be incorporated into a Retina (fundus) camera and the assessment completed real-time. Thus, allowing a point of care assessment quickly and cheaply.