SAIL-VOS 3D
收藏资源简介:
SAIL-VOS 3D数据集是由伊利诺伊大学厄巴纳-香槟分校创建的合成视频数据集,专注于视频数据中的对象检测和3D网格重建。该数据集包含237,611帧,总计3,460,213个对象实例,涵盖178个类别。数据集通过使用GTA-V游戏引擎模拟真实城市环境来生成,提供了丰富的3D结构和多样化的对象类别。SAIL-VOS 3D数据集不仅包含视频帧和相机矩阵,还提供了深度数据、实例级分割、实例级非模态分割以及相应的3D对象形状。此数据集旨在为3D感知算法提供一个良好的测试平台,特别是在处理动态3D形状时,通过开发基准模型来利用时间信息,从而实现更真实和详细的重建。
The SAIL-VOS 3D dataset is a synthetic video dataset created by the University of Illinois Urbana-Champaign, focusing on object detection and 3D mesh reconstruction from video data. This dataset consists of 237,611 frames, with a total of 3,460,213 object instances spanning 178 categories. It is generated using the GTA-V game engine to simulate realistic urban environments, providing rich 3D structures and diverse object categories. In addition to video frames and camera matrices, the SAIL-VOS 3D dataset also offers depth data, instance-level segmentation, instance-level amodal segmentation, and corresponding 3D object shapes. This dataset aims to serve as an excellent testbed for 3D perception algorithms, particularly for dynamic 3D shape processing, by developing benchmark models that leverage temporal information to achieve more realistic and detailed reconstructions.

- 1SAIL-VOS 3D: A Synthetic Dataset and Baselines for Object Detection and 3D Mesh Reconstruction from Video Data伊利诺伊大学厄巴纳-香槟分校 · 2021年



