VEnvision3D

VEnvision3D是由武汉大学国家重点实验室信息工程测量、绘图和遥感创建的大型3D合成感知数据集，旨在支持多任务学习，包括深度完成、分割、上采样、位置识别和3D重建。该数据集利用CARLA模拟器生成，具有在同一环境域内收集的数据，支持多任务模型的探索。数据集的独特之处在于能够在虚拟环境中模拟环境变化和在模型表面上采样点云，为3D视觉领域的多任务模型和基础模型的发展提供了新的挑战和机遇。

VEnvision3D is a large-scale 3D synthetic perceptual dataset developed by the State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University. It aims to support multi-task learning, including depth completion, segmentation, upsampling, place recognition, and 3D reconstruction. Generated using the CARLA simulator, the dataset contains data collected within the same environmental domain, enabling the exploration of multi-task models. The unique feature of this dataset lies in its ability to simulate environmental changes in virtual environments and upsample point clouds on model surfaces, which provides new challenges and opportunities for the development of multi-task models and foundation models in the 3D vision field.