SARA motion (Synthetic Actors and Real Actions)

Sara motion 是一个名为 Synthetic Actors and Real Actions (SARA) 的 3D 运动数据集，用于训练模型以生成适合推理运动相似性的运动嵌入。 该数据集的运动序列数据是通过组合 18 个不同的演员（即执行动作的角色）生成的。使用 Adobe Fuse 软件将角色渲染成骨架形状。选择了四个动作类别（战斗、冒险、运动和舞蹈），包括许多动作变化，其中每个动作的帧长度为 32 或更多。 SARA 数据集中有 4,428 个基本动作（例如，跳舞、跳跃）。Sara Motion is a 3D motion dataset named Synthetic Actors and Real Actions (SARA), which is used to train models to generate motion embeddings suitable for inferring motion similarity. The motion sequence data of this dataset is generated by combining 18 distinct actors, i.e., characters that perform actions. Characters are rendered into skeletal shapes using Adobe Fuse software. Four action categories are selected, namely combat, adventure, sports, and dance, covering a wide range of action variations, with each individual action having a frame length of 32 or greater. The SARA dataset contains 4,428 basic actions (e.g., dancing, jumping).

搜集汇总

背景与挑战

背景概述

SARA motion是一个3D运动数据集，旨在训练模型生成运动嵌入以评估运动相似性。它通过Adobe Fuse软件将18个演员渲染成骨架形状，涵盖战斗、冒险、运动和舞蹈四个类别，包含4,428个基本动作，每个动作帧长度至少为32。