X2C Humanoid Facial-Expression Image-to-Control Dataset
X2C 是人形脸部表情视觉到控制的公开合成训练集：100,000 个虚拟机器人渲染图像与30维连续脸部控制值对。官方 Hugging Face 数据卡为公开非gated、CC-BY-4.0，约4.15GB；官方 MIT 代码提供训练与推理。它可用于表情控制与人类视频到机器人脸部动作研究，但不是人形真机轨迹、操控数据或通用VLA数据。
- 开放状态
- 已核验开放
- 许可证
- CC-BY-4.0 for the dataset card declaration; X2CNet code is separately MIT
- 商业使用
- CC-BY-4.0 attribution obligations apply to the data; code and any target-hardware drivers retain separate terms
- 人工评审结论
- fixed_revision_public_ungated_synthetic_humanoid_expression_dataset
- 机器人
- X2C virtual humanoid face、Ameca used only for qualitative author demonstration
- 任务
- humanoid facial-expression imitation, human-to-humanoid visual expression transfer, and image-to-control regression
- 模态
- rendered humanoid-face RGB image、30-dimensional continuous facial-control vector、dataset metadata
- 格式
- Hugging Face dataset repository; exact file schema and all split manifests were not fully downloaded during review
- 采集/生成方式
- authors render a virtual counterpart of the humanoid face and sample ground-truth control values from animation curves at the rendered-frame timestamps
- 已知问题
- Synthetic rendered facial data only; no full-package checksum or complete local download verification; no released pretrained checkpoint verified; no hardware mapping/calibration, failure ledger, complete split schema or independent reproduction. Do not treat author Ameca media as quantified real-robot evidence.