Aims & Scope
We are entering a new era of Artificial Intelligence (AI), where performance and computational efficiency serve as the two foundational pillars for complex systems comprising hundreds or even thousands of agents. The conventional paradigm, mainly focused on final performance, poses significant challenges, as high-capacity networks drive up costs and result in substantial power consumption due to the vast amounts of data being processed. Addressing these challenges by leveraging tiny, low-power devices is required to reduce dependence on data streaming to powerful servers, enable faster inference, and ensure privacy-preserving processing.
This workshop aims to gather researchers and practitioners to explore these key issues, such as efficient algorithms, enhanced robustness, and the development of resilient systems able to operate effectively in unpredictable environments. Emphasis will be placed on efficient methodologies for model and data optimization, including techniques like distillation, and strategies for improving trustworthiness through explainable methods. It will highlight cutting-edge research and practical solutions that enable the deployment of AI models on resource-constrained, intelligent devices. By providing an inclusive platform, this workshop will facilitate discussions on recent advancements in AI and Artificial Intelligence of Things (AIoT) domains.
Topics
Energy-Efficient Deep Learning Models
Lightweight AI for IoT and AIoT
Model Compression: Pruning, Quantization, and Distillation
Transformers and CNNs Optimization
On-Device Learning
Trustworthy and Explainable AI for IoT Systems
Resilient AI Systems in Unpredictable Environments
Low-Power AI Hardware and Accelerators
Decentralized and Collaborative AI for Smart Devices
Edge Computing for Real-Time AI Inference
AI-driven Predictive Maintenance and Smart Manufacturing
Data-Efficient Learning for Resource-Constrained Devices
Benchmarking AI Efficiency: Trade-offs Between Accuracy and Cost
Neural Architecture Search (NAS) for Edge AI
Details
The SEEDS workshop will be a half-day event. Accepted papers will be included in the ICIP proceedings.
Prospective authors are encouraged to carefully review the following submission requirements before preparing their manuscripts:
Submissions must not exceed 5 pages of technical content, which includes all figures and references. An optional 6th page may be included for references only.
All submissions will be subject to a double-blind peer review process. Please refer to the ICIP main track submission guidelines for further details on the double-blind policy: https://cmsworkshops.com/ICIP2025/ papers/paper_kit.php.
No rebuttal phase will be included in the review process.
Submission URL: https://cmsworkshops.com/ICIP2025/papers/submission.asp?Type=WS&ID=6.
Important Dates
Workshop Paper Submission Deadline: 28 May 2025 4 June 2025
Workshop Paper Acceptance Notification: 25 June 2025 2 July 2025
Workshop Final Paper Submission Deadline: 02 July 2025 9 July 2025
Workshop Author Registration Deadline: 16 July 2025
Workshop Date: 14 September 2025
Program (Tentative)
Each paper presentation will be 15 minutes, followed by 5 minutes for discussion and questions.
Time Event
08:30 – 08:40 Opening remarks and introduction
Papers Session I
08:40 – 09:00 Diffusion based scalable semantic communication framework for image compression and transmission
Namesh Kushantha Pannigala Gamage, Thanuj Fernando, Anil Fernando
09:00 – 09:20 Semantic communication and residual deep learning-driven bitplane coding for scalable image compression and communications
Prabhath Samarathunga, Thanuj Fernando, Yasith Ganearachchi, Nimesh Pollwaththage, Anil Fernando
09:20 – 09:40 Rethinking the backbone in class imbalanced federated source free domain adaptation: the utility of vision foundation models
Kosuke Kihara, Junki Mori, Taiki Miyagawa, Akinori F. Ebihara
09:40 – 10:00 Data efficient stream-based active distillation for scalable edge model deployment
Dani Manjah, Tim Bary, Benoît Gérin, Benoît Macq, Christophe De Vleeschouwer
10:00 – 10:30 Break
Papers Session II
10:30 – 10:50 The low-light-weight network: towards real-time low-light image enhancement for smartphones and other low-power devices
Nikhil Rath, Dinesh Babu Jayagopi, Ramesh Katuri, Pannaga Bhushan Reddy
10:50 – 11:10 MA-YOLO: video object detection via motion-assisted YOLO
Xinyu Wang, Hong-Shuo Chen, Zhiruo Zhou, Jie-En Yao, C.-C. Jay Kuo
11:10 – 11:30 Learning single-image super-resolution in the JPEG compressed domain
Sruthi Srinivasan, Elham Shakibapour, Mehdi Saeedi, Rajy Rawther
11:30 – 11:50 Rethinking time-frequency visualization: towards efficient physiological signal classification
Nidhi Sawant, Swathi Pratapa, Tushar Sandhan
11:50 – 12:00 Closing remarks
Organizers
Pasquale Coscia
Università degli Studi di Milano
Italy
Konstantinos N. Plataniotis
University of Toronto
Canada
Nikolaos Boulgouris
Brunel University of London
United Kingdom
Pai Chet Ng
Singapore Institute of Technology
Singapore
Supported by