[CFP] PRELUDE III Special Session @ IEEE MMSP 2026
Davide Moroni
Please distribute this call to interested parties.
**************************************************
Co-located with:
IEEE MMSP2026 - IEEE International Workshop on Multimedia Signal Processing
Istanbul, Turkey – September 22-24, 2026
Scope
The integration of pervasive sensing infrastructures with distributed computing paradigms has established the foundation for modern smart environments, impacting domains ranging from Industry 4.0 to healthcare. Historically, the distribution of intelligence across the edge-cloud continuum focused on deploying discriminative machine learning models under constrained resources. However, the emergence of Generative Artificial Intelligence (GenAI) and Large Vision-Language Models (VLMs) necessitates a fundamental reassessment of multimedia processing architectures.
The 3rd edition of the PRELUDE special session addresses the theoretical and practical challenges of embedding generative foundation models within heterogeneous, multi-tier networked ecosystems. While cloud-based Artificial Intelligence as a Service (AIaaS) provides the computational capacity for large-scale GenAI, pervasive multimedia understanding requires pushing these capabilities toward the network edge to minimize latency, preserve bandwidth, and ensure data privacy. This session aims to bring to light and discuss the different architectural shifts required to bridge the edge-cloud divide for GenAI, emphasizing distributed inference, model compression, and edge-native foundation models.
Deploying these hybrid architectures in real-world applications (such as industrial automation and public transport monitoring) requires rigorous uncertainty quantification. By establishing calibrated confidence thresholds for edge-based decisions, systems can implement dynamic edge-to-cloud offloading strategies. In this paradigm, the edge handles continuous nominal processing, while highly uncertain or degraded instances (such as complex mobility anomalies) are selectively transmitted to the cloud for advanced generative restoration and analysis. Furthermore, deploying GenAI for continuous multimedia streams introduces severe constraints regarding energy efficiency. A critical objective of this session is to investigate optimal configurations for multimedia processing that align AI deployment with sustainable resource use. Concurrently, GenAI presents novel methodologies for addressing data scarcity in pervasive environments through synthetic data generation and federated learning paradigms.
Topics
Topics of interest include, but are not limited to:
- Architectures for Generative AI and VLMs distributed across the edge-cloud continuum
- Model compression, quantization, and efficient inference of foundation models on constrained systems
- Distributed and federated learning frameworks for generative multimedia models
- Synthetic data generation and generative simulation for scarce data scenarios in multimedia networks
- Energy-efficient computing frameworks and sustainable AI deployment strategies
- Multi-view computer vision and smart camera networks integrated with Vision-Language Models
- Fault-tolerant decision-making and reinforcement learning architectures in heterogeneous sensor networks
- Security, privacy preservation, and adversarial robustness in edge-based generative systems
- Explainability, trustworthiness, and sustainability of pervasive AI paradigms
- Applications in industry, smart cities and smart mobility, cultural heritage and tourism, healthcare, and sustainable agriculture and food systems
Important dates
Submission deadline: May 15, 2026
Acceptance/Reject notification: July 17, 2026
Camera-ready: July 31, 2026
Submission
Each submission should be at most 6 pages in total, including bibliography and well-marked appendices.
Detailed examples of paper formatting are available in the official templates.
The MMSP 2026 paper submission is handled via the Microsoft Conference Management Toolkit (CMT). If you are unfamiliar with CMT you can find instructions here.
All submitted papers will be carefully evaluated based on originality, significance, technical soundness, and clarity of expression by at least two reviewers. The organizers will examine the reviews and make final paper selections.
Publication
All the papers accepted for the special session will be included in the IEEE Xplore Digital Library and major indexes. IEEE MMSP is officially recognized as a CORE B-ranked conference.
Registration
At least one author of each accepted paper must register for the workshop. The workshop registration fee is determined by the organizers.
Signals & Images LAB
Institute of Information Science and Technologies
ISTI-CNR
via Moruzzi 1, 56124, Pisa
e-mail: davide...@isti.cnr.it