metmuseum/openaccess-embeddings-siglip2-naflex

该数据集为metmuseum/openaccess数据集中所有公共领域艺术品的图像嵌入向量，使用google/siglip2-so400m-patch16-naflex模型生成。每个嵌入向量对应一个唯一的objectID，与主数据集中的objectID匹配，并包含L2归一化的1152维浮点数列表、模型来源和嵌入维度信息。数据集不存储图像字节，需通过连接主数据集来获取原始图像。嵌入规范为维度1152，预期图像大小为256像素。

Image embeddings for every public-domain artwork in metmuseum/openaccess, produced by google/siglip2-so400m-patch16-naflex. Each embedding corresponds to an objectID matching the main dataset, with L2-normalised vectors of dimension 1152, model source, and embedding dimension. Image bytes are not stored; join with the main dataset to recover them. Embedding spec: dim=1152, expected image size=256px.