FineMotion
收藏资源简介:
FineMotion数据集是一个包含超过442,000个人类运动片段及其对应详细描述的人体部分运动的数据集。该数据集旨在解决现有文本-运动数据集中描述过于粗糙,缺乏细节的问题。数据集包含约95k个详细段落描述整个人体运动序列的运动。数据集的构建过程包括自动生成和人工标注两个阶段,并且支持零样本细粒度运动编辑,使用户能够通过文本修改来调整运动内容。
The FineMotion Dataset is a human motion dataset comprising over 442,000 human motion clips paired with their corresponding detailed descriptions of partial human body movements. This dataset aims to address the issue that existing text-motion datasets suffer from overly coarse descriptions and a lack of fine-grained details. It includes approximately 95,000 detailed paragraphs that describe full human motion sequences. The construction of the dataset consists of two stages: automatic generation and manual annotation. Moreover, it supports zero-shot fine-grained motion editing, enabling users to adjust motion content via textual modifications.




