STL-10 Dataset 图像识别数据集

STL-10 Dataset 是用于开发无监督特征学习、深度学习、自学习算法的图像识别数据集，其基于 CIFAR-10 Dataset 做了一些修改。其中每个类别均包含比 CIFAR-10 更少的标记训练示例，但提供规模更大的未标记实例用于监督训练的图像学习模型，其利用未标记的数据来构建拥有的先验数据，其中高分辨率图像可被用于开发更具扩展性的无监督学习方法。

The STL-10 Dataset is an image recognition dataset designed for unsupervised feature learning, deep learning, and self-taught learning algorithms, with modifications derived from the CIFAR-10 Dataset. Each category includes fewer labeled training samples than CIFAR-10, while offering a larger corpus of unlabeled instances to support supervised training of image learning models. It leverages unlabeled data to construct prior knowledge, and its high-resolution images can be utilized to develop more scalable unsupervised learning methodologies.