WP4 develops the MARVEL E2F2C framework in order to fully harness, in resource-optimized and secure way, the edge (including data capturing), fog and cloud resources to effectively orchestrate and distribute computational and AI-related tasks. A major consideration towards this goal is the ability to perform a significant part of the processing at the Edge or Fog. Therefore, AI-enabled devices that can run (edge-optimised) light-weight DL models play a key role. Moreover, security and acceleration methods will be enforced at all layers so as to end up with a robust, trustworthy and fast decision support toolkit. To address these challenges, WP4 objectives are to

• offer GPU stream processing acceleration in edge devices and nodes.
• define and deploy a security strategy, including privacy-aware algorithms, at the edge.
• describe innovations performed by analogue and digital microphones that are based on MEMS technology.
• utilize the openSMILE platform for audio analysis and feature extraction.
• develop advanced visualisation techniques to support both real-time and long term decision-making.