3D perception challenge (3D-PC)

3D感知挑战(3D-PC)是由布朗大学脑科学研究所创建的数据集，旨在评估人类和深度神经网络(DNNs)在三维感知任务中的表现。该数据集包含7480张图像，用于训练和测试，涵盖了深度排序和视觉视角获取(VPT)等任务。数据集通过结合3D高斯溅射、视频和Unity技术生成，确保了视觉统计的恒定性和标签的平衡。3D-PC的应用领域包括实时预测人类行为和深化对大脑工作机制的理解，旨在解决机器在三维世界中感知和推理能力的差距。

The 3D Perception Challenge (3D-PC) is a dataset created by the Brain Science Institute at Brown University, designed to evaluate the performance of both humans and deep neural networks (DNNs) in 3D perception tasks. This dataset comprises 7,480 images for training and testing, covering tasks such as depth ordering and Visual Perspective Taking (VPT). Generated by combining 3D Gaussian Splatting, video, and Unity technologies, it ensures consistent visual statistics and balanced labeling. The application scenarios of 3D-PC include real-time prediction of human behavior and deepening the understanding of the brain's working mechanisms, aiming to bridge the gap in perception and reasoning capabilities of machines in the 3D world.