A comprehensive statistic on methods related to deep face restoration.
[Paper]
[Supplementary Material]
💥 Note: More visual comparisons can be found in the Paper and Supplementary Material.
@article{li2023survey,
title={Survey on Deep Face Restoration: From Non-blind to Blind and Beyond},
author={Li, Wenjie and Wang, Mei and Zhang, Kai and Li, Juncheng and Li, Xiaoming and Zhang, Yuhang and Gao, Guangwei and Deng, Weihong and Lin, Chia-Wen},
journal={arXiv preprint arXiv:2309.15490},
year={2023}
}
Pub | Paper | Technology |
---|---|---|
TPAMI2019 | Can We See More? Joint Frontalization and Hallucination of Unaligned Tiny Faces | GAN-based |
FG2020 | IF-GAN: Generative Adversarial Network for Identity Preserving Facial Image Inpainting and Frontalization | GAN-based |
TCSVT2021 | Joint Face Image Restoration and Frontalization for Recognition | GAN/Prior-based |
TCSVT2021 | Simultaneous Face Completion and Frontalization via Mask Guided Two-Stage GAN | GAN/Attention-based |
TIP2021 | Face Hallucination With Finishing Touches | GAN/Prior-based |
Pub | Paper | Technology |
---|---|---|
AAAI2017 | Face Hallucination with Tiny Unaligned Images by Transformative Discriminative Neural Networks | GAN-based |
CVPR2017 | Hallucinating Very Low-Resolution Unaligned and Noisy Face Images by Transformative Discriminative Autoencoders∗ | GAN-based |
CVPR2018 | Semantic Face Hallucination: Super-Resolving Very Low-Resolution Face Images with Supplementary Attributes | GAN-based |
AAAI2020 | Joint Super-Resolution and Alignment of Tiny Faces | Prior-based |
IJCV2020 | Hallucinating Unaligned Face Images by Multiscale Transformative Discriminative Networks | GAN-based |
Pub | Paper | Technology |
---|---|---|
ECCV2018 | Super-Identity Convolutional Neural Network for Face Hallucination | Prior-based |
ICIP2019 | Low-Resolution Face Recognition Based on Identity-Preserved Face Hallucination | CNN-based |
TIP2019 | SiGAN: Siamese Generative Adversarial Network for Identity-Preserving Face Hallucination | GAN/Prior-based |
TIP2019 | Face Hallucination Using Cascaded Super-Resolution and Identity Priors | GAN/Prior-based |
Nercom2019 | Edge and Identity Preserving Network for Face Super-Resolution | GAN/Prior-based |
Arxiv2019 | Verification of Very Low-Resolution Faces Using An Identity-Preserving Deep Face Super-Resolution Network | GAN/Prior-based |
IJCV2019 | Wavelet Domain Generative Adversarial Network for Multi-scale Face Hallucination | GAN/Prior-based |
TPAMI2020 | Adversarial Cross-Spectral Face Completion for NIR-VIS Face Recognition | GAN-based |
Pub | Paper | Technology |
---|---|---|
ICB2019 | SeLENet: A Semi-Supervised Low Light Face Enhancement Method for Mobile Face Unlock | CNN-based |
CVPR2020 | Copy and Paste GAN: Face Hallucination from Shaded Thumbnails | CNN-based |
CVPR2020 | From Less to More: Spectral Splitting and Aggregation Network for Hyperspectral Face Super-Resolution | CNN-based |
ICME2020 | Learning To See Faces In The Dark | CNN-based |
Arxiv2021 | Network Architecture Search for Face Enhancement | CNN-based |
TPAMI2021 | Recursive Copy and Paste GAN: Face Hallucination From Shaded Thumbnails | GAN-based |
TMM2023 | Learning to Hallucinate Face in the Dark | CNN-based |
AAAI24 | Low-Light Face Super-resolution via Illumination, Structure, and Texture Associated Representation | Attention-based |
Pub | Paper | Technology |
---|---|---|
ICIP2021 | Identity and Attribute Preserving Thumbnail Upscaling | Prior-based |
ICML2021 | Fairness for image generation with uncertain sensitive attributes | CNN-based |
TPAMI2022 | EDFace-Celeb-1M: Benchmarking Face Hallucination With a Million-Scale Dataset | Dataset |
Arxiv2022 | Blind Face Restoration: Benchmark Datasets and a Baseline Model | Dataset |
Pub | Paper | Technology |
---|---|---|
FG202O | Face Denoising and 3D Reconstruction from A Single Depth Image | CNN-based |
IJCB2021 | 3D Face Point Cloud Super-Resolution Network | CNN-based |
Sensors2022 | Incomplete Region Estimation and Restoration of 3D Point Cloud Human Face Datasets | CNN-based |
CVPR2022 | Learning to Restore 3D Face from In-the-Wild Degraded Images | Prior/GAN-based |
💥 Note: It can be found in the Supplementary Material.
Dataset | Quantity | Introduction | Year |
---|---|---|---|
FFHQ | 70,000 | non-paired dataset for training | 2018 |
CelebA | 202,599 | non-paired dataset for training | 2015 |
CelebA-HQ | 30,000 | non-paired dataset for training | 2020 |
LFW | 13,233 | non-paired dataset for training | 2008 |
Multi-PIE | 75,000 | non-paired dataset for training | 2010 |
Helen | 2,330 | non-paired dataset for testing | 2012 |
CelebA-Test | 3,000 | non-paired dataset for testing | 2021 |
CelebChild-Test | 180 | non-paired real-world dataset for testing | 2021 |
CelebAdult-Test | 180 | non-paired real-world dataset for testing | 2021 |
WebPhoto-Test | 407 | non-paired real-world dataset for testing | 2021 |
LFW-Test | 1,711 | non-paired real-world dataset for testing | 2021 |
Wider-Test | 970 | non-paired real-world dataset for testing | 2022 |
💥 Note: More datasets can be found in the Paper.
🚩 Note: A detailed description of the evaluation metrics and how to use it can be found here.
If you have any question, please email lewj2408@gmail.com