Syn-Mediverse

Syn-Mediverse是首个针对医疗设施智能场景理解的多模态合成数据集，由德国弗莱堡大学计算机科学系创建。该数据集包含超过48,000张模拟工业标准光学跟踪相机捕获的图像，并提供超过150万个注释，涵盖深度估计、物体检测、语义分割、实例分割和全景分割等五种场景理解任务。数据集创建过程中，使用NVIDIA Isaac Sim模拟器，通过多相机设置捕捉图像，并在不同光照条件下从13个不同的医疗房间中收集数据。Syn-Mediverse的应用领域包括提高手术效率、辅助手术审计和提升医疗机器人的自主性，旨在解决医疗环境中复杂场景理解的问题。

Syn-Mediverse is the first multimodal synthetic dataset tailored for intelligent scene understanding in healthcare facility scenarios, developed by the Department of Computer Science, University of Freiburg, Germany. This dataset includes over 48,000 images captured by simulated industry-standard optical tracking cameras, with more than 1.5 million annotations covering five scene understanding tasks: depth estimation, object detection, semantic segmentation, instance segmentation, and panoptic segmentation. During the dataset construction, the NVIDIA Isaac Sim simulator was utilized to capture images via a multi-camera configuration, with data collected from 13 distinct medical rooms under varying lighting conditions. The application domains of Syn-Mediverse include improving surgical efficiency, assisting surgical auditing, and enhancing the autonomy of medical robots, aiming to address the challenges of complex scene understanding in healthcare environments.