Express4D
收藏资源简介:
Express4D数据集是动态面部表情生成的关键数据集,包含丰富的面部运动序列和语义标注。数据集由18名参与者表演,涵盖了1205个多样化的序列。数据集利用LLM生成的自然语言指令,并以流行的ARKit blendshape格式进行采集,易于使用普通设备收集,并与行业标准动画流程兼容。该数据集旨在解决现有面部运动数据集缺乏细粒度控制和复杂表情表达的局限性,通过自然语言驱动的动态面部表情生成,为动画、虚拟角色和人类计算机交互等领域提供丰富的资源。
The Express4D dataset is a pivotal resource for dynamic facial expression generation, boasting abundant facial motion sequences and semantic annotations. Collected from performances by 18 participants, it encompasses 1205 diverse sequences. The dataset utilizes LLM-generated natural language instructions and is captured in the widely adopted ARKit blendshape format. It facilitates collection with consumer-grade devices and is compatible with industry-standard animation workflows. This dataset aims to address the limitations of existing facial motion datasets, namely the lack of fine-grained control and support for complex facial expression rendering, and provides abundant resources for domains including animation, virtual characters, and human-computer interaction through natural language-driven dynamic facial expression generation.

- 1Express4D: Expressive, Friendly, and Extensible 4D Facial Motion Generation Benchmark特拉维夫大学 · 2025年



