Efficient Machine Learning
This use case develops a deep learning optimiser that uses OmpSs to reduce the energy consumption of deep learning inference significantly. The optimiser make use of OmpSs to increase the computational architectures supported by the optimiser to CPU, GPU and FPGA. The focus and aim is to facilitate users to deploy deep learning based AI on embedded systems in an energy- and cost-efficient manner.
This use case searches for a new method to discover biomarker candidates in datasets with a very low number of samples and up to thousands of features. For this research a very high computational power is needed. The LEGaTO technology with the usage of heterogeneous hardware provides the possibility to enter new areas of research combined with a reduction of energy consumption in bioinformatics.
Urban-scale air quality model
In many urban areas air quality and associated impacts on public health are matters of growing concern. The emission and dispersion of critical pollutants (PMx, O2 and ground-level O3) correlate with cancer, asthma, cardiorespiratory problems, brain development in children and reduction of life expectancy in general. As a consequence, air quality monitoring networks and modelling forecasting systems are critical to increase awareness and, ultimately, to assist decision-makers on the adoption of measures to protect public health.
In this scenario, the LEGaTO stack will be pivotal, both in terms of leveraging the processing capabilities and improving the energy-efficiency of an operational urban-scale air quality modelling system. The Smart City use case aims at demonstrating that monitoring of urban air quality through CFD simulations is feasible for nowcasting predictions in an operational workflow.
Smart Mirror / Human Interaction Interface
Intelligent and interconnected technology is becoming more and more popular in all areas of life. In our homes, heating, blinds and light are often already being operated by speech control. However, these algorithms are often too complex and require too much computing power for energy-efficient usage. Within LEGaTO, an energy-optimized processing platform is developed to tackle this challenge.
For a central interaction interface of a smart apartment the smart mirror is developed. Using depth-image cameras, it recognizes users and displays personalised information, including, for example, individualised bus schedules, the menu of the day at the cafeteria, or current information about the apartment. To handle this variety of applications, both hand gesture control and voice input can be used. The implemented algorithms are based on machine learning methods. The high-performance embedded hardware platform allows for data – some of which may be personal – to be processed locally. This negates the need for data to be processed on external servers at a third-party company, ensuring that privacy is maintained.
This use case leverages the following LEGaTO open-source contribution: SmartMirror GitHub
Secure IoT Gateway
Secure IoT Gateway
When it comes to Industry 4.0 and IoT, the communication of the local devices among each other using local or wide area networks can be associated with a high level of complexity. Especially securing the network connections of IoT is difficult. The Secure IoT Gateway will simplify the complexity of securing the connection of devices to a network, coming with a Network Cockpit Application for configuring and monitoring the system.
The Secure IoT Gateway is a special use case of the LEGaTO project as its general goal is not energy efficiency but to simplify the complexity of communication of local devices to a network. So it helps other use cases to achieve their goals by reducing the complexity of security for them.