Scribble

该数据集名为Scribble，包含了十个物体类别（篮球、鸡、饼干、纸杯蛋糕、月亮、橙子、菠萝、足球、草莓和西瓜）的图像与简笔画配对，所有图像均具有统一的白色背景。由于大量物体共享相似的圆形特征，该数据集对生成算法提出了挑战，其中60%的物体类别几乎具有完全相同的圆形形状，这为算法带来了显著的模糊性。该数据集的规模为1512对图像与简笔画（1412对用于训练，100对用于测试），其任务是进行简笔画到图像的翻译。

The dataset is named Scribble. It contains image and scribble pairs for ten object categories: basketball, chicken, biscuit, cupcake, moon, orange, pineapple, football, strawberry, and watermelon. All images have a uniform white background. Since many objects share similar circular features, this dataset poses challenges for generative algorithms; 60% of the object categories have nearly perfectly identical circular shapes, introducing significant ambiguity for the algorithms. The dataset consists of 1512 image-scribble pairs in total, with 1412 pairs used for training and 100 pairs for testing. The task of this dataset is scribble-to-image translation.