RenderMe-360

RenFace是一个大规模多视角人脸高清视频数据集，包含多样的人脸表情、丰富的细粒度发型发色，同时也包含音素均衡的说话视频。可应用于：2D/3D数字人脸生成、人脸重建捕捉等领域。

亮点:
- 大规模数据（ID>500，视频帧>80M）
- 多样的表情（自然表情+11种夸张表情）
- 多样的发型发色 (自然发型/发色+各类假发+头套)
- 音素均衡的话术
- 多样的视角（60个）
- 高清的图像（2k）
- 丰富的标注 (前背景分割、相机参数、2D/3D人脸关键点、FLAME模型拟合系数、人头稠密重建、外观与显著特征、文本动作描述、AU)

RenFace is a large-scale multi-view high-definition facial video dataset that encompasses diverse facial expressions, rich fine-grained hairstyles and hair colors, as well as phonetically balanced speaking videos. It can be applied to fields such as 2D/3D digital face generation, face reconstruction and capture, etc.

Highlights:
- Large-scale data: over 500 unique identities and more than 80 million video frames
- Diverse expressions: including natural expressions plus 11 types of exaggerated expressions
- Diverse hairstyles and hair colors: covering natural hairstyles/hair colors, various wigs and head covers
- Phonetically balanced speech scripts
- Diverse viewpoints: 60 distinct viewing angles
- High-definition resolution: 2K
- Rich annotations: including foreground-background segmentation, camera parameters, 2D/3D facial landmarks, FLAME model fitting coefficients, dense head reconstruction, appearance and salient features, text-based action descriptions, and Action Units (AU)