RetouchingFFHQ
收藏资源简介:
RetouchingFFHQ是由复旦大学创建的大规模人脸修饰数据集,包含652,568张修饰图像和58,158张未修饰图像。数据集通过Megvii、Alibaba和Tencent的在线API进行修饰,涵盖四种常见的修饰操作:眼睛放大、面部提升、皮肤平滑和面部美白,每种操作分为四个级别:关闭、轻微、中等和重度。该数据集用于训练和测试人脸修饰检测模型,特别适用于细粒度检测和多修饰类型识别,旨在解决社交媒体平台上的真实性问题和欺骗性广告的影响。
RetouchingFFHQ is a large-scale face retouching dataset created by Fudan University. It contains 652,568 retouched images and 58,158 unretouched images. The dataset is generated via the online APIs of Megvii, Alibaba and Tencent, covering four common retouching operations: eye enlargement, face lifting, skin smoothing and face whitening, each of which has four levels: off, mild, moderate and severe. This dataset is used for training and testing face retouching detection models, and is particularly suitable for fine-grained detection and multi-retouching-type recognition, aiming to address the authenticity issues and the impact of deceptive advertisements on social media platforms.

- 1RetouchingFFHQ: A Large-scale Dataset for Fine-grained Face Retouching Detection复旦大学 · 2023年



