official press release are two well-known Japanese company – a manufacturer of spare parts and components DENSO Corporation and Toshiba Corporation electronics developer – announced an agreement to jointly develop artificial intelligence for the car. The joint project is called «Deep Neural Network-Intellectual Property» (DNN-IP). The final development will include self-developed by both companies technology to recognize objects that will help create a system of driver assistance and lead to automatic driving technologies.
As the name of the project, object recognition systems Car technology is based on the simulation of the human brain works. This algorithm is based on in-depth training of the neural network. In the future system based on Deep Neural Network technology will have to perform better – faster and more accurate than it is capable person
The modern machine learning system based on the analysis of a car cameras images, taking into account the previously uploaded images. It is obvious that in this case, to consider all possible options for the traffic situation is simply impossible. deep learning systems based on neural networks are able to learn from the data generated by the system during the scanning space. Thus, the list of identified objects is expanding rapidly, and the increasing recognition accuracy.
Partners in the face of companies DENSO and Toshiba plan to create a whole range of solutions for deep learning machines using neural networks. learning unit may be so small that it can be incorporated into the processor for controlling the vehicle in automotive systems or surveillance camera.
Integrated system to assist the driver in the car or autopilot control will develop the company of DENSO . Toshiba The company takes on the task to implement the technology “smart” recognition of objects on the road in the form of electronic circuits. The effectiveness of specialized solutions DNN-IP promises to be higher than the widely used in the field of general-purpose digital signal processors or graphics cards.