AI Paper Atlas

Research map / Benchmarks / CelebA

CelebA leaderboard

225 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on CelebA. Best FID: 2.97 (CPSample (α=0.001, 0.05), ICLR 2025).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

CelebA · FID

118 papers, each with its best reported number for its own method; lower is better.

#FID ↓MethodPaperTopic
12.97CPSample (α=0.001, 0.05)
DDIM
CPSample: Classifier Protected Sampling for Guarding Training Data During Diffusion
ICLR 2025 · Table: FID Score comparison
Fast diffusion sampling
23.21BlackOut (Ours)Blackout Diffusion: Generative Diffusion Models in Discrete-State Spaces
ICML 2023 · Table 2
Fast diffusion sampling
33.51DDIM (S=1000, η=0.0)Denoising Diffusion Implicit Models
ICLR 2021 · Table 1
Fast diffusion sampling
43.7FPA + U-Net
FPA
Images as Weight Matrices: Sequential Image Generation Through Synaptic Learning Rules
ICLR 2023 · Table 1
Autoregressive and latent generation
54CT-StyleGAN2Exploiting Chain Rule and Bayes' Theorem to Compare Probability Distributions
NeurIPS 2021 · Table 3
Variational and generative sampling
64.15CoopFlow(Pre)A Tale of Two Flows: Cooperative Learning of Langevin Flow and Normalizing Flow Toward Energy-Based Model
ICLR 2022 · Table: FID scores on CIFAR-10,
Variational and generative sampling
74.69SSCGANSSCGAN: Facial Attribute Editing via Style Skip Connections
ECCV 2020 · Table 5
Editing and style transfer
85.21GEBM (ULA)
SNGAN ConvNet
Generalized Energy Based Models
ICLR 2021 · Table 2
Variational and generative sampling
95.36MC-DiT-B / 2
DiT-B / 2
MC-DiT: Contextual Enhancement via Clean-to-Clean Reconstruction for Masked Diffusion Models
NeurIPS 2024 · Table 9
Fast diffusion sampling
105.38Our KD
Hierarchical VAE, 0.44M parameters
A Unified Knowledge Distillation Framework for Deep Directed Graphical Models
CVPR 2023 · Table 1
Variational and generative sampling
108 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

CelebA · worst-group accuracy

51 papers, each with its best reported number for its own method; higher is better.

41 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

CelebA · accuracy

38 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
199.07CNN-DCTLeveraging Frequency Analysis for Deep Fake Image Recognition
ICML 2020 · Source identification table
GANs
295.9FairNet-Partial
ViT
FairNet: Dynamic Fairness Correction without Performance Loss via Contrastive Conditional LoRA
NeurIPS 2025 · Table 1
Recognition and long-tail
395.16CAGlow
Glow
Conditional Adversarial Generative Flow for Controllable Image Synthesis
CVPR 2019 · Table 1
Autoregressive and latent generation
495.1Positive LS (α=0.1)
ResNet-152
Be Careful What You Smooth For: Label Smoothing Can Be a Privacy Shield but Also a Catalyst for Model Inversion Attacks
ICLR 2024 · Table 1
Privacy attacks and secure learning
594.31DCS²
ConvNet
Concealing Sensitive Samples against Gradient Leakage in Federated Learning
AAAI 2024 · Table 2
Privacy attacks and secure learning
694ExtGraft
ResNet-18 (ImageNet pretrained)
Recover Fair Deep Classification Models via Altering Pre-trained Structure
ECCV 2022 · Table 3
Pruning and efficient tuning
791.8BISE + finetuningBias In, Bias Out? Finding Unbiased Subnetworks in Vanilla Models
CVPR 2026 · Table 5
Recognition and long-tail
891.8FADES
ResNet-18
FADES: Fair Disentanglement with Sensitive Relevance
CVPR 2024 · Table 1
Recognition and long-tail
991.62LearnToBranch-Deep-Wide
GNAS-Deep-Wide
Learning to Branch for Multi-Task Learning
ICML 2020 · CelebA results table
RNNs and normalization
1091.55FedXDS (LRP)
ResNet8
FedXDS: Leveraging Model Attribution Methods to counteract Data Heterogeneity in Federated Learning
ICCV 2025 · Table 3
Personalized and heterogeneous FL
28 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

CelebA · PSNR

25 papers, each with its best reported number for its own method; higher is better.

15 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

CelebA · SSIM

23 papers, each with its best reported number for its own method; higher is better.

13 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

CelebA · LPIPS

19 papers, each with its best reported number for its own method; lower is better.

9 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

CelebA · average accuracy

12 papers, each with its best reported number for its own method; higher is better.

2 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.