A Real World Dataset for Multi-view 3D Reconstruction
收藏资源简介:
本数据集名为'A Real World Dataset for Multi-view 3D Reconstruction',由西蒙弗雷泽大学和阿里巴巴XR实验室联合创建。数据集包含998个日常桌面物品的3D模型及其847,000张真实世界RGB和深度图像。每张图像均通过半自动化方式进行了精确的相机姿态和物体姿态标注,以支持多种3D应用,如形状重建、物体姿态估计和形状检索等。数据集主要关注基于学习的多种视角3D重建,旨在填补缺乏适当真实世界基准的空白。数据集内容丰富,包括高分辨率、纹理化的3D模型,以及详细的图像和姿态信息。创建过程中,使用了专业的3D扫描设备和视频录制技术,确保了数据的高质量。数据集的应用领域广泛,主要用于推动3D物体理解领域的技术进步。
This dataset, named 'A Real World Dataset for Multi-view 3D Reconstruction', was jointly created by Simon Fraser University and Alibaba XR Lab. It contains 3D models of 998 everyday desktop objects, along with 847,000 real-world RGB and depth images. Each image has been accurately annotated with camera poses and object poses via a semi-automatic workflow, supporting a variety of 3D applications including shape reconstruction, object pose estimation, shape retrieval, and more. The dataset primarily focuses on learning-based multi-view 3D reconstruction, aiming to fill the gap caused by the lack of appropriate real-world benchmarks. It features rich content, including high-resolution, textured 3D models as well as detailed image and pose annotation information. During its development, professional 3D scanning equipment and video recording technologies were employed to ensure the high quality of the dataset. With a wide range of application scenarios, this dataset is mainly used to drive technological advancements in the field of 3D object understanding.

- 1A Real World Dataset for Multi-view 3D Reconstruction西蒙弗雷泽大学 · 2022年



