Aloha project

Software framework for runtime-Adaptive and secure deep Learning On Heterogeneous Architectures

Horizon 2020

This project is funded by the European Union’s Horizon 2020 programme, grant agreement No. 780788

Consortium

We are a team of 14 partners from 7 countries

Deep Learning (DL) algorithms are an extremely promising instrument in artificial intelligence. To foster their adoption in new applications and markets, a step forward is needed towards the implementation of DL inference on low-power embedded systems, enabling a shift to the edge computing paradigm. The main goal of ALOHA is to facilitate implementation of DL algorithms on heterogeneous low-energy computing platforms providing automation for optimal algorithm selection, resource allocation and deployment.

ALOHA Use cases

Speech recognition in smart industry

This scenario refers to Smart Industry, where Deep Learning is used for speech recognition. The objective of this use case is to develop an embedded speech recognition system that would activate/deactivate PLC-controlled tooling machinery or collaborative robot in an industrial environment, without relying on a cloud backend.


Read more and see the videos...

Surveillance of Critical Infrastructures

This scenario refers to Critical Infrastructures surveillance, where Deep Learning is meant to be used to define an intelligent video-based detection system for security in and around the critical infrastructure. A specific challenge is to use deep learning methods to enable technician after a short training to utilize advanced capabilities of modern surveillance techniques.

Read more...

Medical decision assistant

This use-case refers to a DL-based smart assistant, which supports emergency room situations, identifying acute intracranial bleeds in non-contrast CT images. The consortium wants to assess the benefits provided by the ALOHA tool flow within the development of an embedded medical decision assistant, considering the requirements posed by the specific application in terms of accuracy and performance.

Read more...

Architecture-awareness

The features of the architecture that will execute the inference are taken into account during the whole development process, starting from the early stages such as pre-training hyperparameter optimization and algorithm configuration.

Productivity

The tool flow implements support for agile development methodologies, to be easily adopted by SMEs and midcaps.

Adaptivity

The development process considers that the system should adapt to different operating modes at runtime.



Extensibility

The development process is conceived to support novel processing platforms to be exploitable beyond the end of the project.

Security

The development process automates the introduction of algorithm features and programming techniques improving the resilience of the system to attacks.


Parsimonious inference

In DL, good precision levels can be also obtained using algorithms with reduced complexity. All the optimization utilities in the ALOHA tool flow will consider the effects of algorithm simplification on precision, execution time, energy and power.

Latest News & Events

ALOHA Final Workshop - June 30, 2021 - Teams

The ALOHA Final Workshop has been held virtually on June 30th, 2021 through the online meeting platform Teams at: https://rebrand.ly/aloha-final-workshop

In this event, the ALOHA consortium presented the results achieved within the project, the assessment of the toolflow capabilities on the reference design cases, and a summary of the main technical solutions and research outcomes. The workshop was open to researchers and developers in the ICT community, from academy and industry, involved in both hardware and software development, as well as companies (including SMEs) interested in a smoother transition to deep learning for their products. Participation to the workshop was free of charge.

Read more...

 

Buffer Sizes Reduction for Memory-efficient CNN Inference on Mobile and Embedded Devices - Video presentation

Convolutional Neural Networks (CNNs) are the core of many intelligent systems, including those that run on mobile and embedded devices. The execution of computationally demanding and memory-hungry CNNs on resource-limited mobile and embedded devices is quite challenging. One of the main problems,when running CNNs on such devices, is the limited amountof memory available...

In this video Svetlana Minakova and Todor Stefanov, from the Leiden Institute of Advanced Computer Science, present the paper entitled "Buffer Sizes Reduction for Memory-efficient CNN Inference on Mobile and Embedded Devices" which was accepted at the Euromicro Conference on Digital System Design 2020.

Read more...

Combining task- and data-level parallelism for high-throughput CNN inference on embedded CPUs-GPUs MPSoCs - Video presentation

Nowadays Convolutional Neural Networks (CNNs) are widely used to perform various tasks in areas such as computer vision or natural language processing. Some of the CNN applications require high-throughput execution of the CNN inference, on embedded devices...

In this video Svetlana Minakova, Erqian Tang and Todor Stefanov, from the Leiden Institute of Advanced Computer Science, present the paper entitled "Combining task- and data-level parallelism for high-throughput CNN inference on embedded CPUs-GPUs MPSoCs" which was accepted at the SAMOS XX International Conference on Embedded Computer Systems: Architectures, Modeling and Simulation.

Read more...

Contacts

Project Coordinator
Giuseppe Desoli - STMicroelectronics
giuseppe(dot)desoli(at)st(dot)com

Scientific Coordinator
Paolo Meloni - University of Cagliari, EOLAB
paolo(dot)meloni(at)diee(dot)unica(dot)it

Dissemination Manager
Francesca Palumbo - University of Sassari, IDEA Lab
fpalumbo(at)uniss(dot)it

Twitter

Linkedin