AI-POWERED MIXED REALITY

AI and Deep Learning are profoundly changing the way in which Mixed Reality applications perceive their environment. Indeed, the addition of intelligence to applications will contribute to providing an experience that is truly engaging and increasingly closer to the concept of Intelligent Edge, with data analysed locally, where it is generated, thus minimising latency timeframes.

Valorem Reply is working in the Intelligent Edge realm with interesting results, such as those obtained by the application of AI on a HoloLens device. The video shows a demo app that uses a Deep Learning model trained to recognise specific objects. The model was incorporated into an app for HoloLens, exploiting Windows Machine Learning (WinML) in order to enable the recognition of objects in near real time. Thanks to the HoloLens’ spatial mesh, it is also possible to detect the distance of the objects, which is transferred to the user by speech synthesis.

Over the next few years we will see a rapid growth of AI acceleration chips for Intelligent Edge, an innovation that is fuelling the widespread use of fast Machine Learning models on low latency customer devices. These VPU/NPU (vision processing unit/neural processing unit) acceleration chips will soon be available in all devices, just like GPUs (graphics processing unit). The use of these chips, together with intelligent cloud solutions, will enable the development of interesting use cases capable of creating value in several areas.

In Valorem Reply, we are already exploiting the potential offered by these worlds, during the phases associated with the creation of a Machine Learning model. Cloud models are trained to harness the power of cloud computing, together with the ability to scale resources during the most intense training phase. Subsequently, the Deep Learning algorithm is evaluated on edge devices with higher latency, so as to provide results in real time.