Introduction

With the growth of surveillance applications, the volume of multimedia surveillance data—spanning acoustic signal, radio, image, video, text, physiological signal, and metadata—has become increasingly large-scale, heterogeneous, and multimodal. Manual monitoring is not always practical, which calls for intelligent, efficient, trustworthy, and automated surveillance solutions. Firstly, massive amounts of multimedia surveillance data in storage necessitate advanced analysis techniques such as event detection, action and activity recognition, anomaly detection, and cross-modal summarization. These capabilities are critical for applications like public safety monitoring, transportation management, healthcare assessment, smart cities and buildings, and industrial inspection. Secondly, semantic-level information (e.g., object trajectories, bounding boxes, 3D human poses, scene graphs, and textual annotations) extracted from multimedia analytics has become essential for intelligent surveillance systems. New challenges are emerging in semantic data representation, efficient storage and transmission, cross-scene generalization, and privacy-preserving semantic processing. Thirdly, the evolution from centralized computing toward collaborative and distributed computing environments—including edge–cloud architectures, federated learning, and sensor-network cooperation—poses new research questions in multi-sensor fusion, distributed signal processing, collaborative perception, and trustworthy & explainable decision making. Fourthly, surveillance data often involves sensitive information, raising concerns about data security, user privacy, and ethical usage. Research is increasingly focusing on developing privacy-preserving analytics, secure multi-party computation, federated learning approaches, and robust methods that are resilient to adversarial attacks, sensor failures, or missing data. Ensuring fairness, accountability, and transparency in surveillance processing does not only enhance system reliability but also builds public trust, which is critical for widespread adoption in real-world applications. This workshop aims to provide an international forum for researchers, engineers, and practitioners to present the latest innovations and share experiences on the design, modeling, deployment, and evaluation of advanced surveillance data processing technologies.

This is the 7th edition of our workshop. The first six were organized as the BIG-Surv workshop in conjunction with ICME 2019 (Shanghai, China), ICME 2020 (London, UK), ICME 2021 (Shenzhen, China), ICME 2022 (Taipei, Taiwan ROC), ICME 2023 (Brisbane, Australia) and ICME 2025 (Nantes, France). This year, the BIG-Surv workshop has been renamed as SDP with a new scope and focus.