Playing Cards

该数据集名为“玩牌”，是一个合成图像数据集，旨在分析具有完美概念注释的概念基础模型（CBMs）。它包含40,000张图像，分为四种变体：单一玩牌、三张随机玩牌、用于分类手牌等级的三张玩牌，以及简化概念的扑克牌版本。每种变体都有70%的训练图像和30%的验证图像，这样的比例分配旨在让数据集在分析时不受噪声干扰。这个数据集的规模为40,000张图像，其任务是评估那些具有清晰输入特征映射到概念标签的CBMs。This dataset, named "Card Playing", is a synthetic image dataset developed for analyzing Concept-Based Models (CBMs) with perfect concept annotations. It contains 40,000 images, divided into four variants: single playing card, three random playing cards, three playing cards for hand rank classification, and a simplified playing card version. For each variant, 70% of the images are allocated to the training set and 30% to the validation set. Such a split ratio is designed to ensure that the dataset is free from noise during analysis. Overall, this 40,000-image dataset is intended to evaluate CBMs that exhibit clear mappings between input features and concept labels.

搜集汇总

背景与挑战

背景概述

该数据集是一个包含四组扑克牌图像的数据集，每组有10,000张图像，总计规模在10K到100K之间，适用于图像分类和图像分割任务，特别是多标签分类和实例分割。数据集中图像经过随机旋转、翻转和缩放处理，背景随机，并提供了概念标签（如花色和点数）、任务分类标签（如扑克手牌等级）以及坐标信息，旨在用于测试概念瓶颈模型。