ryushinn/CelebA-Filtered

--- dataset_info: features: - name: image dtype: image - name: bbox list: int32 length: 4 - name: facial_landmarks struct: - name: lefteye list: int32 length: 2 - name: righteye list: int32 length: 2 - name: nose list: int32 length: 2 - name: leftmouth list: int32 length: 2 - name: rightmouth list: int32 length: 2 splits: - name: train num_bytes: 1414170037 num_examples: 45843 - name: test num_bytes: 148017298 num_examples: 7242 - name: validation num_bytes: 145114402 num_examples: 6786 download_size: 1708445921 dataset_size: 1707301737 configs: - config_name: default data_files: - split: train path: data/train-* - split: test path: data/test-* - split: validation path: data/validation-* --- CelebA in the wild images, bounding boxes, and facial landmarkds. Images where the face covers less than 30% of the area are filtered out, resulting in * Train size: 45843 * Validation size: 6786 * Test size: 7242 Original dataset: https://mmlab.ie.cuhk.edu.hk/projects/CelebA.html Processed from the original Huggingface dataset: https://huggingface.co/datasets/hfaus/CelebA_bbox_and_facepoints