ManiCups

该数据集是一个关于州级别图像操作的集合，它要求模型通过对容器进行填充或清空液体来操作杯子，被设计为一个用于对象状态变化的多领域图像到图像翻译数据集。该数据集包含5个领域，其中有3个丰富领域（空杯、咖啡、果汁），每个类别都有超过1000张图像，以及2个低资源领域（水、牛奶），每个类别不到1000张图像。整个数据集大约有5700张图像，包括空杯以及装有咖啡、果汁、牛奶和水的杯子。这项任务的目的是进行多领域图像到图像翻译，以应对对象状态的变化。

This dataset is a collection of state-level image manipulation tasks that require models to manipulate cups by filling or emptying liquids in containers. It is designed as a multi-domain image-to-image translation dataset targeting object state changes. The dataset includes 5 domains, among which 3 are high-resource domains (empty cup, coffee, juice) with over 1000 images per category, and 2 low-resource domains (water, milk) with fewer than 1000 images per category. The entire dataset has approximately 5700 images in total, covering empty cups and cups filled with coffee, juice, milk and water. The purpose of this task is to perform multi-domain image-to-image translation to handle object state variations.