HyperFace
收藏资源简介:
HyperFace数据集是由Idiap研究所、洛桑联邦理工学院和洛桑大学共同创建的合成面部识别数据集。该数据集包含10,000个合成面部图像,旨在解决传统面部识别数据集中的隐私和伦理问题。数据集的创建过程通过在预训练面部识别模型的嵌入空间中进行优化,生成具有高类间和类内变化的合成面部图像。HyperFace数据集主要应用于面部识别模型的训练,旨在提升合成数据集在面部识别任务中的表现。
The HyperFace dataset is a synthetic face recognition dataset jointly created by Idiap Research Institute, École Polytechnique Fédérale de Lausanne and University of Lausanne. This dataset contains 10,000 synthetic face images, and is designed to address the privacy and ethical issues inherent in traditional face recognition datasets. The dataset is constructed by optimizing within the embedding space of a pre-trained face recognition model, generating synthetic face images with high inter-class and intra-class variations. The HyperFace dataset is primarily applied to the training of face recognition models, aiming to improve the performance of synthetic datasets on face recognition tasks.

- 1HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding HypersphereIdiap研究所, 洛桑联邦理工学院, 洛桑大学 · 2024年



