This project aims at providing a platform for the rapid prototyping of software applications based on intelligent video analysis (IVA) for commercial activity with table service (i.e., restaurants). The platform is named VIDIO, and its purpose is to facilitate the application of Artificial Intelligence in the context of food&service to support the personnel service, training, and safety. To satisfy actual GDPR rules and to guarantee real-time behaviors, IVA applications have to be executed “at-the-edge”, whereby sensitive information kept by cameras is elaborated close to the input sensor by low-power heterogeneous computing devices. The VIDIO platform aims at supporting the rapid prototyping of such IVA applications through a model-based design flow, by which software applications are modelled, integrated, and validated at high level of abstraction and, then, semi-automatically synthesized on a target device by taking into consideration both functional and non-functional constraints (i.e., performance, power consumption, energy efficiency). The project starts from an existing prototype built for the development of Robotics applications for Industry 4.0. It will extend the existing platform to support the development of heterogeneous SW applications, such as Deep Learning inference and data mining beside image processing for embedded systems. To tackle the intrinsic complexity of such a heterogeneity, different development languages will be combined to the de-fact modelling standard for embedded vision applications (i.e., OpenVX). Intelligent algorithms and heuristics will be defined to provide efficient mapping and scheduling of application tasks over the heterogenous computing elements of the target device, with the aim of guaranteeing the overall embedded system real-time performance and energy efficiency.
