athenlab/reva

--- dataset_info: features: - name: image dtype: image - name: keywords sequence: string - name: caption dtype: string - name: generated_image dtype: image splits: - name: train configs: - config_name: default data_files: - split: train path: data/* license: apache-2.0 --- reva is a dataset of real images and their "equivalent" ai-generated images. the real images are first captioned with moondream, then the caption is used to generate the corresponding ai image. reva is primarily made for training an ai image detection model.