aklein4/coyo-llava-hq

该数据集包含多个特征字段，包括URL、不同类型的标题、标签、关键字段、CLIP相似度评分、NSFW评分、水印评分、美学评分、人脸数量、图像尺寸、EXIF信息和SHA256哈希值。数据集分为训练集、验证集和测试集，分别包含不同数量的样本和字节大小。This dataset contains multiple feature fields, including URL, different types of captions (caption_llava, caption_llava_short, caption), tags (tags_open_images, tags_booru), key field, CLIP similarity scores (clip_similarity_vitb32, clip_similarity_vitl14), NSFW scores (nsfw_score_opennsfw2, nsfw_score_gantman), watermark score, aesthetic score (aesthetic_score_laion_v2), number of faces (num_faces), image dimensions (width, height), EXIF information, and SHA256 hash. The dataset is divided into training, validation, and test sets, each containing a different number of samples and byte sizes.