Skip to content

Latest commit

 

History

History
32 lines (21 loc) · 1.45 KB

datasets.md

File metadata and controls

32 lines (21 loc) · 1.45 KB

Datasets

CelebA-HQ

Brief description

CelebFaces Attributes HQ Dataset (CelebA-HQ) is a large-scale face attributes dataset that consists of 30,000 high-resolution face images selected from the CelebA dataset, each of which annotated according to the following 40 attributes:

no_beard, young, attractive, mouth_slightly_open, smiling, wearing_lipstick, high_cheekbones, male, heavy_makeup, wavy_hair, oval_face, pointy_nose, arched_eyebrows, big_lips, black_hair, big_Nose, brown_hair, straight_hair, bags_under_eyes, wearing_earrings, bangs, blond_hair, bushy_eyebrows, narrow_eyes, wearing_necklace, 5_o_clock_shadow, receding_hairline, wearing_necktie, rosy_cheeks, eyeglasses, goatee, chubby, sideburns, blurry, wearing_hat, double_chin, pale_skin, gray_hair, mustache, bald

Download

Download the dataset following the instructions given here (e.g., following this link in Google Drive). The root directory of the dataset (dataset_root) should have the following structure:

├── annotations
│   ├── CelebA-HQ-to-CelebA-mapping.txt
│   ├── CelebAMask-HQ-attribute-anno.txt
│   └── list_eval_partition.txt
└── data
    ├── 0.jpg
    ├── 1.jpg
    ├── ...
    ├── 9998.jpg
    └── 9999.jpg