Artificial vision system powered by deep neural networks.
The problem we solve
Conventional optical inspection systems suffer from serious limitations when working in a changing environment. Any minimal modification in the lighting conditions, a shadow, a reflection, remains of dirt, or a simple human fingerprint, can lead to a incorrect rejection for the element to be verified, increasing the production costs and reducing the efficiency and quality ratios.
This limitation creates a nasty compromise that needs to be faced by the companies trying to deploy classic artificial vision systems in their production chain: the higher the applied quality standard is for the visual inspection to be executed, the higher is the number of parts that are erroneously rejected too.
The target for GL Research when creating the Neurona Vision solution is to solve this and other related problems that affect to the conventional artificial vision solutions by applying the most advanzed artificial intelligence technology.
The solution we propose
Neurona Vision is a complete solution including all the hardware and software elements required to build any kind of vision and artificial intelligence system. In this way, Neurona Vision allows us to deploy in a record time from compact systems formed by a single device or brain and targeted to isolated machines, to complex artificial neural systems built upon multiple devices and targeted to provide global intelligence at a factory scale level.
At the software side, Neurona Vision includes from a high performance operating system integrating several tools and utilities, to a fully reconfigurable user interface designed for the control of artificial vision and intelligence systems and translated to multiple languages.
About the hardware, the Neurona Vision portfolio is composed by several families of application specific electronic equipment that are totally compatibles with each other and that are able to be integrated with the products from the main providers of actuators, sensors, cameras and robots in the market.
Our technological advantage
In the conventional artificial vision systems, the human is the one in charge of studying the measurements and tests to be performed over a image to later explain this procedure to the machine by using a rather complex mathematical and programming language. With the Neurona Vision systems from GL Research, the human only needs to verify and classify the images and then the machine deduces in an autonomous way the internal criteria to be applied so that its decisions are identical to those made by the human who train it.
This learning capability in the Neurona Vision systems, that can be progressively perfected over the time while the solution is serving on the field, is based in the intensive use of different deep neural network models. Employing this kind of artificial neural networks requires a extreme computational power to be commercially feasible, and here is when the technological advantage from GL Research is manifested in the Neurona Vision solution.
Our extensive experience in the design of high performance chips at the micro and nanometric scale, joined to our close collaboration with the main electronic technology providers in the world, not only allow us to totally optimize our artificial intelligence software for its execution using the most advanced processing systems in the market, but we have been able to build our own chips capable of accelerating artificial vision and intelligence tasks and targeted to those scpecific applications when an extra performance and efficiency boost is required.
In this way, the Neurona Vision devices from GL Research are able to process thousands of images per seconds when connected to the power electrical grid, such as those integrated in an industrial production line, or hundreds of images per second when powered by batteries, such as those boarded as payload in AGVs, drones or other autonomous service vehicles.