It's this time of the year that most people have come back from their summer holidays, they are recharged and full of creativity! Here in MARVEL, we have one more reason to feel energized and inspired: We just had our mid-term project review! Thanks to the hard work of the whole consortium and the good collaboration among the partners the feedback was both rewarding for all the progress achieved so far, and constructive for what is further expected.

With this Newsletter, we would like to proudly share what happened in the last 4 months as well as what is coming up! Enjoy and don't forget to follow us online (if you are not doing it already), for more immediate updates!

MARVEL and AI Hub Tampere jointly organized the “Audio and Speech Technology Workshop” on 16th June at Paidia workspace in Nokia Arena, Tampere, Finland.
The development of machine learning in combination with signal processing techniques has led to new possibilities in applications that process and analyze audio or speech. The workshop aimed to provide an opportunity to exchange information about the latest development in these fields.

Visit our zenodo community and get access to a great number of scientific papers, presentations, and deliverable reports. The outcomes of our latest research include:

• PhiNets: a scalable backbone for low-power AI at the edge
• Gradient Based Clustering
• Personalized Federated Learning via Convex Clustering
• Federated Feature Selection for Cyber-Physical Systems of Systems,
• Stateless or stateful FaaS? I'll take both!
• On the Role of Smart Vision Sensors in Energy-Efficient Computer Vision at the Edge
• Single-Layer Transformers for More Accurate Early Exits with Less Overhead
• Dynamic Split Computing for Efficient Deep Edge Intelligence
• Automatic Social Distance Estimation From Images: Performance Evaluation, Test Benchmark, and Algorithm
• Weighted Edit Distance for Country Code Recognition in License Plates
• On Efficiently Partitioning a Topic in Apache Kafka
• Evaluating Short-Term Forecasting of Multiple Time Series in IoT Environments
• Optimizing PhiNet architectures for the detection of urban sounds on low-end devices

By the end of June 2022 (M18 of the project), our consortium prepared 10 deliverables 7 of which are publicly available and are enlisted below:

• D2.2: Management and distribution Toolkit – initial version | Access at zenodo here
• D3.1: Multimodal and privacy-aware audio-visual intelligence – initial version | Access at zenodo here
• D3.2: Efficient deployment of AI-optimised ML/DL models – initial version | Access at zenodo here
• D4.2: Security assurance and acceleration in E2F2C framework – initial version | Access at zenodo here
• D4.3: MARVEL's decision-making toolkit – initial version | Access at zenodo here
• D5.3: HPC infrastructure and resource management for audio-visual data analytics – initial version | Access at zenodo here
• D6.1: Demonstrators execution - initial version | Access at zenodo here

Click on the button below and view all our deliverables!

An outline of upcoming activities