CUB-200-2011 (Caltech-UCSD Birds-200-2011)

Caltech-UCSD Birds-200-2011 (CUB-200-2011) 数据集是用于细粒度视觉分类任务的最广泛使用的数据集。它包含属于鸟类的 200 个子类别的 11,788 张图像，其中 5,994 张用于训练，5,794 张用于测试。每个图像都有详细的注释：1 个子类别标签、15 个部分位置、312 个二进制属性和 1 个边界框。文本信息来自 Reed 等人。他们通过收集细粒度的自然语言描述来扩展 CUB-200-2011 数据集。为每个图像收集十个单句描述。自然语言描述是通过 Amazon Mechanical Turk (AMT) 平台收集的，要求至少 10 个单词，没有任何子类别和动作信息。

Caltech-UCSD Birds-200-2011 (CUB-200-2011) dataset is one of the most widely utilized datasets for fine-grained visual classification tasks. It comprises 11,788 images across 200 bird subcategories, with 5,994 images allocated for training and 5,794 for testing. Each image is accompanied by comprehensive annotations: 1 subcategory label, 15 part locations, 312 binary attributes, and 1 bounding box. The textual annotations were derived from Reed et al., who extended the CUB-200-2011 dataset by collecting fine-grained natural language descriptions. Ten single-sentence descriptions were gathered for each image. These natural language descriptions were collected via the Amazon Mechanical Turk (AMT) platform, with the constraints that each description must contain at least 10 words and must not include any subcategory or action-related information.