VSD2M
收藏资源简介:
VSD2M是由微信人工智能和腾讯创建的目前最大的视觉语言贴纸数据集,包含209万条静态和动态贴纸数据。该数据集通过网络爬取、数据过滤、标注和数据集分割四个阶段构建,涵盖了丰富的情感和动作描述,特别适用于多帧动画贴纸生成任务。数据集的创建旨在解决动画贴纸生成领域的数据获取困难和基准不完善的问题,推动智能创作领域的研究。
VSD2M is currently the largest visual-language sticker dataset developed by WeChat AI and Tencent. It consists of 2.09 million static and dynamic sticker samples. This dataset is constructed through four stages: web crawling, data filtering, annotation, and dataset splitting. It covers rich emotional and action-related descriptions, and is particularly well-suited for multi-frame animated sticker generation tasks. The dataset was created to address the challenges of limited data availability and incomplete benchmarks in the animated sticker generation domain, and to advance research in the field of intelligent creation.




