</> Code with Papers by Syntology

Datasets › FFHQ

FFHQ (Flickr-Faces-HQ)

Introduced by Tero Karras et al. in A Style-Based Generator Architecture for Generative Adversarial Networks1 Jan 2019 archive 2025-07-28

Flickr-Faces-HQ (FFHQ) consists of 70,000 high-quality PNG images at 1024×1024 resolution and contains considerable variation in terms of age, ethnicity and image background. It also has good coverage of accessories such as eyeglasses, sunglasses, hats, etc. The images were crawled from Flickr, thus inheriting all the biases of that website, and automatically aligned and cropped using dlib. Only images under permissive licenses were collected. Various automatic filters were used to prune the set, and finally Amazon Mechanical Turk was used to remove the occasional statues, paintings, or photos of photos.

Source: Flickr-Faces-HQ Dataset (FFHQ) Image Source: https://github.com/NVlabs/ffhq-dataset

HomepageBenchmarksPapers JSON

Benchmarks archive 2025-07-28

All 17 leaderboards whose dataset resolves to this page shown (sort by any header). "First row" is the archive's own first row at snapshot, in the archive's row order; nothing here re-ranks and metric direction is not asserted.

First row (archive order)PaperCode
Image Generation FFHQ 256 x 256 StyleSAN-XL FID 1.68 SAN: Inducing Metrizability of GAN with Discriminative... sony/san 51 Compare
Image Generation FFHQ 1024 x 1024 StyleSAN-XL FID 1.61 SAN: Inducing Metrizability of GAN with Discriminative... sony/san 20 Compare
Image Generation FFHQ-U Alias-Free-R FID 3.66 Alias-Free Generative Adversarial Networks NVlabs/stylegan3 +6 13 Compare
Image Generation FFHQ Anycost GAN FID 2.77 Anycost GANs for Interactive Image Synthesis and Editing mit-han-lab/anycost-gan 12 Compare
Image Super-Resolution FFHQ 256 x 256 - 4x upscaling HiFaceGAN FID 5.36 HiFaceGAN: Face Renovation via Collaborative Suppression... Lotayou/Face-Renovation +4 11 Compare
Image Super-Resolution FFHQ 1024 x 1024 - 4x upscaling HiFaceGAN FID 1.978 HiFaceGAN: Face Renovation via Collaborative Suppression... Lotayou/Face-Renovation +4 9 Compare
Image Super-Resolution FFHQ 512 x 512 - 4x upscaling HiFaceGAN PSNR 30.824 HiFaceGAN: Face Renovation via Collaborative Suppression... Lotayou/Face-Renovation +4 8 Compare
3D-Aware Image Synthesis FFHQ 256 x 256 CIPS-3D FID 6.97 CIPS-3D: A 3D-Aware Generator of GANs Based on... PeterouZh/CIPS-3D 4 Compare
Face Hallucination FFHQ 512 x 512 - 16x upscaling HiFaceGAN FID 11.389 HiFaceGAN: Face Renovation via Collaborative Suppression... Lotayou/Face-Renovation +4 4 Compare
Image Generation FFHQ 512 x 512 StyleSAN-XL FID 1.77 SAN: Inducing Metrizability of GAN with Discriminative... sony/san 3 Compare
Image Inpainting FFHQ 512 x 512 SH-GAN FID 3.4 Image Completion with Heterogeneously Filtered Spectral Hints shi-labs/sh-gan 3 Compare
Image Inpainting FFHQ 1024 x 1024 BRGM LPIPS 0.19 Bayesian Image Reconstruction using Deep Generative Models razvanmarinescu/brgm +1 2 Compare
3D-Aware Image Synthesis FFHQ 512 x 512 - 4x upscaling IDE-3D FID 4.6 IDE-3D: Interactive Disentangled Editing for... mrtornado24/ide-3d 1 Compare
Facial Inpainting FFHQ DMFN SSIM 0.8985 Image Fine-grained Inpainting Zheng222/DMFN +2 1 Compare
Image Denoising FFHQ BRGM LPIPS 0.24 Bayesian Image Reconstruction using Deep Generative Models razvanmarinescu/brgm +1 1 Compare
Image Denoising FFHQ 64x64 - 4x upscaling BRGM LPIPS 0.24 Bayesian Image Reconstruction using Deep Generative Models razvanmarinescu/brgm +1 1 Compare
Image Generation FFHQ 64x64 - 4x upscaling PFGM++ FID 2.43 PFGM++: Unlocking the Potential of Physics-Inspired... newbeeer/pfgmpp 1 Compare

Papers archive 2025-07-28

30 shown of 59 papers with a leaderboard row on this dataset's benchmarks, newest first. The archive's own "papers using this dataset" list was never published, so this is the benchmark-backed subset; the archive's count for this dataset is 1,468. The Syntology column is from Syntology's graph (read 2026-09-25), stated per sample; it is not part of any archive number.

DateSamples run Syntology
The GAN is dead; long live the GAN! A Modern GAN Baseline 1 1 9 Jan 2025 ran 9 of 11 samples (2 unverified; 11 pointer-only for licence)
CLR-GAN: Improving GANs Stability and Quality via Consistent Latent Representation and Reconstruction 1 1 30 Sep 2024 not harvested
Compensation Sampling for Improved Convergence in Diffusion Models 1 1 11 Dec 2023 not harvested
Flow Matching in Latent Space 1 1 17 Jul 2023 ran 1 of 1 samples (0 unverified; 1 pointer-only for licence)
Exposing flaws of generative model evaluation metrics and their unfair treatment of diffusion models 4 9 7 Jun 2023 ran 30 of 63 samples (33 unverified; 1 pointer-only for licence)
Polynomial Implicit Neural Representations For Large Diverse Datasets 1 1 20 Mar 2023 not harvested
PFGM++: Unlocking the Potential of Physics-Inspired Generative Models 1 1 8 Feb 2023 ran 2 of 3 samples (1 unverified; 3 pointer-only for licence)
SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layer 1 3 30 Jan 2023 ran 2 of 2 samples (0 unverified; 2 pointer-only for licence)
StyleNAT: Giving Each Head a New Perspective 2 3 10 Nov 2022 ran 1 of 1 samples (0 unverified)
Image Completion with Heterogeneously Filtered Spectral Hints 1 1 7 Nov 2022 not harvested
Diffusion-GAN: Training GANs with Diffusion 4 1 5 Jun 2022 ran 14 of 18 samples (4 unverified)
IDE-3D: Interactive Disentangled Editing for High-Resolution 3D-aware Portrait Synthesis 1 1 31 May 2022 not harvested
Efficient-VDVAE: Less is more 1 3 25 Mar 2022 not harvested
Polarity Sampling: Quality and Diversity Control of Pre-Trained Generative Networks via Singular Values 1 1 3 Mar 2022 not harvested
StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasets 2 4 1 Feb 2022 ran 14 of 19 samples (5 unverified)
StyleSwin: Transformer-based GAN for High-resolution Image Generation 1 4 20 Dec 2021 ran 5 of 5 samples (0 unverified; 3 pointer-only for licence)
Compositional Transformers for Scene Generation 1 1 1 Dec 2021 not harvested
Unleashing Transformers: Parallel Token Prediction with Discrete Absorbing Diffusion for Fast High-Resolution Image Generation from Vector-Quantized Codes 3 2 24 Nov 2021 not harvested
Projected GANs Converge Faster 3 1 1 Nov 2021 ran 38 of 49 samples (11 unverified; 6 pointer-only for licence)
CIPS-3D: A 3D-Aware Generator of GANs Based on Conditionally-Independent Pixel Synthesis 1 4 19 Oct 2021 ran 6 of 10 samples (4 unverified)
MaGNET: Uniform Sampling from Deep Generative Network Manifolds Without Retraining 1 1 15 Oct 2021 not harvested
Alias-Free Generative Adversarial Networks 7 15 23 Jun 2021 ran 25 of 31 samples (6 unverified; 9 pointer-only for licence)
Improved Transformer for High-Resolution GANs 1 5 14 Jun 2021 ran 3 of 4 samples (1 unverified)
Soft Truncation: A Universal Training Technique of Score-based Diffusion Model for High Precision Score Estimation 1 1 10 Jun 2021 ran 4 of 12 samples (8 unverified)
Data-Efficient Instance Generation from Instance Discrimination 1 2 8 Jun 2021 ran 1 of 1 samples (0 unverified; 1 pointer-only for licence)
Gotta Go Fast When Generating Data with Score-Based Models 1 2 28 May 2021 ran 2 of 3 samples (1 unverified; 3 pointer-only for licence)
Regularizing Generative Adversarial Networks under Limited Data 1 1 7 Apr 2021 ran 8 of 9 samples (1 unverified; 3 pointer-only for licence)
Bridging Global Context Interactions for High-Fidelity Image Completion 1 1 2 Apr 2021 not harvested
Large Scale Image Completion via Co-Modulated Generative Adversarial Networks 1 1 18 Mar 2021 not harvested
Anycost GANs for Interactive Image Synthesis and Editing 1 3 4 Mar 2021 ran 0 of 1 samples (1 unverified)

The full list of 59 is in the JSON twin.

Dataset loaders archive 2025-07-28

docs.activeloop.ai/datasets/ffhq-datasetactiveloopai/Hubtfpytorch
NVlabs/ffhq-dataset

2 loaders as listed in the archive; links are outbound and not re-checked here.

Tasks archive 2025-07-28

Image GenerationImage Super-ResolutionImage DenoisingImage InpaintingFacial Inpainting3D-Aware Image SynthesisFace Hallucination

License archive 2025-07-28

CC BY-NC-SA 4.0

Modalities archive 2025-07-28

Images

Languages archive 2025-07-28

No language tagged.

Variants archive 2025-07-28

  • FFHQ 64x64 - 4x upscaling
  • FFHQ 512 x 512
  • FFHQ-U
  • FFHQ 512 x 512 - 4x upscaling
  • FFHQ 512 x 512 - 16x upscaling
  • FFHQ 256 x 256 - 4x upscaling
  • FFHQ 256 x 256
  • FFHQ 1024 x 1024 - 4x upscaling
  • FFHQ 1024 x 1024
  • FFHQ

10 variant names, as the archive lists them.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections