aipicasso/megalith-10m-florence2

Megalith-10M是一个类似于CC-0的图像数据集，但原始数据集不包含图像标题。因此，使用Florence-2模型为这些图像添加了标题。该数据集适用于视觉与语言研究，以及开发文本到图像或图像到文本模型。添加标题的过程使用了Florence-2-large模型，并花费了3000 GPU小时（T4）。

Megalith-10M is an image dataset similar to CC-0, but the original dataset does not contain image captions. To supplement this, captions were generated for the images using the Florence-2 model. The dataset is primarily used for vision and language research, as well as developing text-to-image or image-to-text models.