D4D Dataset
收藏资源简介:
D4D数据集是由德国国家肿瘤疾病中心与德累斯顿工业大学等机构联合创建,专为评估非刚性腹部手术场景的4D重建而设计。该数据集包含超过30万帧内窥镜视频和369个点云,数据来源于六只猪的腹部手术场景,使用达芬奇Xi立体内窥镜和Zivid结构光相机采集。数据集通过光学追踪和手动迭代对齐方法进行注册,包含三种序列类型:整体变形、增量变形和移动相机片段,旨在测试算法对非刚性运动、变形幅度和视野外更新的鲁棒性。数据集主要应用于非刚性SLAM、4D重建和深度估计方法的开发和评估,解决腹部手术中软组织变形重建的难题。
The D4D dataset was jointly created by institutions including the German National Center for Tumor Diseases and Technische Universität Dresden, and is purpose-built for evaluating 4D reconstruction in non-rigid abdominal surgical scenarios. This dataset contains over 300,000 frames of endoscopic video and 369 point clouds, with data collected from abdominal surgical procedures performed on six pigs using the da Vinci Xi stereoscopic endoscope and Zivid structured-light camera. The dataset is registered via optical tracking and manual iterative alignment methods, and encompasses three sequence types: global deformation, incremental deformation, and moving camera clips, which are designed to test algorithm robustness against non-rigid motion, varying deformation magnitudes, and out-of-view updates. It is primarily utilized for the development and evaluation of non-rigid SLAM, 4D reconstruction and depth estimation methods, addressing the challenge of soft tissue deformation reconstruction in abdominal surgery.

- 1The Dresden Dataset for 4D Reconstruction of Non-Rigid Abdominal Surgical Scenes德国国家肿瘤疾病中心·转化外科肿瘤学系; 德累斯顿工业大学医学院; 德累斯顿工业大学·触觉互联网中心; 香港中文大学·计算机科学与工程系 · 2026年



