novel motion datasets
收藏资源简介:
该数据集由京都大学的研究团队创建,旨在模拟自然环境中物体的运动,特别是具有不同材料属性的物体。数据集包含两种版本:一种是纯Lambertian(哑光)物体,另一种是经历光学湍流的非Lambertian物体。数据集通过训练模型来估计物体运动,并验证模型是否能够感知二阶运动。数据集的应用领域包括计算机视觉和认知科学,旨在解决机器视觉模型与人类视觉系统在运动感知上的差异问题。
This dataset was created by a research team at Kyoto University, aiming to simulate the motion of objects in natural environments, particularly those with different material properties. The dataset includes two variants: one composed of purely Lambertian (matte) objects, and the other featuring non-Lambertian objects that undergo optical turbulence. It is utilized to train models for object motion estimation and validate whether models can perceive second-order motion. Its application areas cover computer vision and cognitive science, with the objective of resolving the discrepancies in motion perception between machine vision models and the human visual system.

- 1Machine Learning Modeling for Multi-order Human Visual Motion Processing京都大学信息学研究科 · 2025年



