• Learning-Based Multi-View Stereo, MVSNet and Its Variants, Depth Estimation, and 3D/4D Reconstruction (基于学习的多视图立体视觉、深度估计、点云分割与模型重建)
• Neural Radiance Fields, 3D/4D Gaussian Splatting, Neural Rendering, Novel View Synthesis, and Dynamic Scene Modeling (神经辐射场、三维/四维高斯泼溅、神经渲染、新视角合成与动态场景建模)
• Domain-Specific Foundation Models, Digital-Twin World Models, Physical AI, and Embodied Autonomous Systems for Real-World Perception, Reasoning, Decision-Making, and Action (面向真实世界感知、推理、决策与行动的领域大模型、数字孪生世界模型、物理人工智能及具身自主系统)
• Computational Imaging, Cross-Modal Alignment, Multimodal Data Fusion, and Joint Image–Signal–Geometry Representation (计算成像、跨模态对齐、多模态数据融合与图像—信号—几何联合表征)
• Spatiotemporal Data Mining, State Estimation, Change and Anomaly Detection, and Predictive Maintenance (时空数据挖掘、状态估计、变化与异常检测及预测性维护)
• Physics-Informed Learning, Physical Modeling, Data Assimilation, and Continuous Physical–Virtual Synchronization (物理信息学习、物理机理建模、数据同化与虚实持续同步)
• Real-Time Updating, Uncertainty Quantification, Fidelity Assessment, and Trustworthy Validation of Digital Twins (数字孪生的实时更新、不确定性量化、保真度评估与可信验证)
• Digital-Twin Applications in Smart Manufacturing, Intelligent Infrastructure, Healthcare, Robotics, Cultural Heritage, and Smart Environments (数字孪生在智能制造、智慧基础设施、医疗健康、机器人、文化遗产与智慧环境中的应用)

This track focuses on digital-twin technologies driven by computer vision, image processing, and signal processing. It investigates how multimodal data—including images, videos, 3D geometry, signals, and operational time series—can be integrated with spatiotemporal data mining, domain-specific foundation models, and world models to enable high-fidelity 3D/4D reconstruction, state perception, evolution prediction, continuous updating, and trustworthy validation of physical objects, environments, and processes, while supporting embodied autonomous systems that perceive, reason, decide, and act in the real physical world. Topics include learning-based multi-view stereo, Neural Radiance Fields (NeRF), 3D/4D Gaussian Splatting (3DGS/4DGS), dynamic scene modeling, computational imaging, cross-modal alignment and data fusion, change and anomaly detection, Physical AI and embodied autonomous systems, physics-informed learning, data assimilation, and physical–virtual synchronization. Particular emphasis is placed on accuracy, robustness, real-time performance, uncertainty quantification, and verifiable fidelity in complex real-world environments. The track welcomes contributions on theory, algorithms, datasets, benchmarks, systems, and applications in smart manufacturing, intelligent infrastructure, healthcare, robotics, cultural heritage, and smart environments.