AnimeCeleb
收藏资源简介:
AnimeCeleb是一个大规模的动画头部数据集,专为动画头部重演设计。该数据集由韩国科学技术院创建,包含240万张高质量图像及其对应的姿态向量。数据集通过利用3D动画模型作为可控图像采样器,能够提供大量头部图像及其详细的姿态注释。为了简化数据创建过程,开发了一个基于Blender的半自动管道和一个开发的注释系统。AnimeCeleb数据集的应用领域包括动画头部重演和直观的姿态编辑,旨在解决动画领域的头部重演问题,并推动相关研究的发展。
AnimeCeleb is a large-scale anime head dataset specifically designed for anime head reenactment. Developed by the Korea Advanced Institute of Science and Technology (KAIST), this dataset contains 2.4 million high-quality images along with their corresponding pose vectors. By leveraging 3D animation models as controllable image samplers, it enables the generation of abundant head images with detailed pose annotations. To streamline the data creation workflow, a Blender-based semi-automatic pipeline and a dedicated annotation system were developed. The application domains of the AnimeCeleb dataset include anime head reenactment and intuitive pose editing, aiming to address the head reenactment challenges in the animation field and advance relevant research.
AnimeCeleb 数据集概述
数据集介绍
- 名称: AnimeCeleb
- 全称: Large-Scale Animation CelebHeads Dataset for Head Reenactment
- 发布机构: Kangyeol Kim, Sunghyun Park, Jaeseong Lee, Sunghyo Chung, Junsoo Lee, Jaegul Choo
- 机构详情:
- KAIST
- Korea University
- Naver Webtoon
- Letsur Inc.
- 发布会议: ECCV 2022
数据集内容
- 目的: 用于动画头部重演(head reenactment)
- 特点:
- 使用3D动画模型作为可控图像采样器,提供大量头部图像及其对应的详细姿态标注。
- 构建了一个半自动化的数据创建流程,利用开放的3D计算机图形软件和开发的标注系统。
- 应用:
- 训练头部重演模型,产生高质量的动画头部重演结果。
- 提出了一种新的姿态映射方法和架构,用于跨域头部重演任务。
数据集下载
相关论文
- 论文标题: AnimeCeleb: Large-Scale Animation CelebHeads Dataset for Head Reenactment
- 论文链接: arXiv
引用
@inproceedings{kim2021animeceleb, title={AnimeCeleb: Large-Scale Animation CelebHeads Dataset for Head Reenactment}, author={Kim, Kangyeol and Park, Sunghyun and Lee, Jaeseong and Chung, Sunghyo and Lee, Junsoo and Choo, Jaegul}, booktitle={Proc. of the European Conference on Computer Vision (ECCV)}, year={2022} }




