Ffhq Dataset | Awesome Repository

This project provides a high-resolution face dataset consisting of 70,000 human face images in PNG format. It serves as a curated library of aligned images and facial landmark data designed for generative model training, facial recognition, and image synthesis research.

The dataset includes machine-readable metadata that pairs images with precise facial coordinate points, source URLs, and copyright information. This coordinate data enables the transformation of raw photos into a standardized 1024x1024 pixel resolution through landmark-based alignment and cropping.

The repository includes automation tools for asset retrieval, featuring a downloader that utilizes concurrent network connections and checksum verification to ensure data integrity. It also provides capabilities for image inclusion verification and general facial image preprocessing.

Features

Face Datasets - Provides a high-quality collection of 70,000 human face images for training generative models.
Facial Landmark Datasets - Provides precise facial coordinate points mapped to images in machine-readable JSON format.
Image-Text Pair Mappings - Pairs high-resolution images with JSON metadata containing facial coordinates and source information.
Training Data Generation - Provides a diverse, large-scale image set curated for deep learning and synthetic image generation.

Features

Face Datasets - Provides a high-quality collection of 70,000 human face images for training generative models.
Facial Landmark Datasets - Provides precise facial coordinate points mapped to images in machine-readable JSON format.
Image-Text Pair Mappings - Pairs high-resolution images with JSON metadata containing facial coordinates and source information.
Training Data Generation - Provides a diverse, large-scale image set curated for deep learning and synthetic image generation.