Sign in to view source links and access this dataset
Description
30,000 high-resolution face images selected from the CelebA dataset, following the CelebA-HQ selection process. Each image includes a corresponding segmentation mask for facial attributes. The dataset, created by author cpuimage, was last updated on March 1, 2025.
Use Cases
Train hair segmentation models based on the provided 1024x1024 auto-annotated masks.
Evaluate hair matting algorithms using the high-resolution face images and masks.
Develop facial attribute parsing models leveraging the segmentation masks corresponding to CelebA attributes.
Strengths
Contains 30,000 high-resolution face images, providing a substantial scale for training.
Masks are auto-annotated at a resolution of 1024 x 1024, offering detailed segmentation targets.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment for specific training needs.
Provenance
Source
Extended from the CelebAMask-HQ dataset, which is derived from the CelebA dataset.
Collection Method
Images were selected by following the CelebA-HQ process; masks were auto-annotated.
Time Range
null
Freshness
Last updated 2025-03-01 17:05:45; freshness should be verified.
Geography
null
License is unknown; terms of use must be verified before application.