LAION-400M

The 'LAION-400M' dataset contains a large-scale collection of images and their associated textual descriptions, and it is used to evaluate image retrieval methods by clustering these images into pseudo classes based on joint textual and visual features.