VinAIResearch/EFHQ

Code and data for the CVPR24 paper "EFHQ: Multi-purpose ExtremePose-Face-HQ dataset" [CVPR'24]

23
/ 100
Experimental

Training deep learning models for facial analysis often struggles with faces viewed from extreme angles. This project provides EFHQ, a high-quality dataset of up to 450,000 faces in various extreme poses, curated from existing high-resolution video datasets. Researchers and developers working on facial synthesis, generation, or recognition can use this data to significantly improve model performance and generalization across diverse head orientations.

No commits in the last 6 months.

Use this if your deep learning models for facial analysis perform poorly when encountering faces viewed from the side, top, or other non-frontal angles.

Not ideal if you are looking for a dataset primarily focused on frontal-view faces or for general image classification tasks unrelated to human faces.

facial-recognition computer-vision-research deep-learning-datasets generative-AI-faces facial-synthesis
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 7 / 25
Maturity 16 / 25
Community 0 / 25

How are scores calculated?

Stars

29

Forks

Language

Python

License

AGPL-3.0

Last pushed

Jul 23, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/VinAIResearch/EFHQ"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.