Tutorial creators: UNISS and UNICA
Researchers from the University of Cagliari (UNICA) and the University of Sassari (UNISS) delivered a tutorial on Adaptive CNN Execution on Edge FPGAs, showcasing their advancements in AI support at the edge. The tutorial, developed by Francesco Ratto, Federico Manca, Claudio Rubattu, and Francesca Palumbo, demonstrates how the MDC tool has been extended to generate adaptive inference engines for Convolutional Neural Networks (CNNs).
By developing efficient and adaptive inference engines for edge FPGAs, UNICA and UNISS actively contribute to the objectives of the MYRTUS project, which aims to advance AI-enabled continuum computing across heterogeneous platforms, optimizing performance and adaptability for real-world applications.
This tutorials was selected for presentation at the prestigious 42nd IEEE International Conference on Computer Design (ICCD 2024), held in November in Milan. ICCD is a leading conference in the field of computer design, covering topics such as computer architecture, digital systems, reconfigurable computing, embedded systems, and hardware-software co-design. It brings together researchers and industry experts to discuss cutting-edge advancements in computing and design automation. [19.11.2024 – IEEE ICCD 2024]
Prior to its selection at ICCD, the tutorial was also presented at two high-profile summer schools:
- The 5th Summer School on Cyber-Physical Systems and the Internet of Things, held in June in Budva [22.06.2024, SS-CPSIoT’2024]
- The 6th CPS Summer School, held in September in Alghero [20.09.2024 , CPS Summer School]
These events provided an excellent platform for engaging with researchers and students, emphasizing the importance of efficient AI execution.
The tutorial is publicly available on GitHub at https://github.com/mdc-suite/qonnx2mdc, allowing researchers and practitioners to explore and implement its methodologies.
