EMMa

EMMa是由斯坦福大学开发的一个大规模、多模态、多任务的对象数据集，包含超过280万个亚马逊产品列表，每个对象都附有图像、列表文本、质量、价格、产品评级和在亚马逊产品类别分类法中的位置等信息。数据集还设计了一个包含182种物理材料的全面分类法，并利用智能标注框架快速为所有对象添加新的二进制标签，减少了手动标注的工作量。EMMa数据集适用于计算机视觉和自然语言处理中的多任务学习，能够帮助解决机器学习实践中生产标注数据集的特定需求问题。

EMMa is a large-scale, multimodal, multi-task object dataset developed by Stanford University. It contains over 2.8 million Amazon product listings, with each object accompanied by information such as images, listing texts, product quality, prices, product ratings, and its position within the Amazon product category taxonomy. The dataset also incorporates a comprehensive taxonomy covering 182 types of physical materials, and utilizes an intelligent annotation framework to quickly add new binary labels to all objects, reducing the workload of manual annotation. EMMa is applicable to multi-task learning in computer vision and natural language processing, and can help address the specific needs of production-grade annotation datasets in machine learning practice.