极市平台(ExtremeMart)是深圳极视角旗下的专业视觉算法开发与分发平台,为开发者提供行业场景集,每月上百真实项目需求,算法分发,技术共享等,旨在联合开发者建立起良好的计算机视觉生态。已与上百名开发者建立了合作并转化了上百种视觉算法。
PS.本周四(11月15日)晚,TEE首席架构师、TEE AI Lab资深研究员邓文彬将为我们讲解如何在GPU/CPU/移动端高效训练CNN网络,公众号回复“35”即可获取直播详情。
图像生成一直是计算机视觉领域非常有意思的方向,图像到图像的变换是其中一个非常重要的应用,使用图像到图像的变换,可以完成非常多有趣的应用,可以把黑熊变成熊猫,把你的照片换成别人的表情,还可以把普通的照片变成毕加索风格的油画,自从GAN横空出世之后,这方面的应用也越来越多,下面是对这个领域的相关论文的一个整理,而且大部分都有代码!
github地址:https://github.com/ExtremeMart/image-to-image-papers
这是一个图像到图像的论文的汇总。
论文按照arXiv上第一次提交时间排序。
监督学习
Note | Model | Paper | Conference | paper link(arXiv) | code link(github) |
---|
|
pix2pix | Image-to-Image Translation with Conditional Adversarial Networks | CVPR 2017 | 1611.07004 | junyanz/pytorch-CycleGAN-and-pix2pix
|
| Contextual GAN | Image Generation from Sketch Constraint Using Contextual GAN | ECCV 2018 | 1711.08972 |
|
| pix2pix-HD | High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs | CVPR 2018 | 1711.11585 | NVIDIA/pix2pixHD |
一对多 | BicycleGAN | Toward Multimodal Image-to-Image Translation | NIPS 2017 | 1711.11586 | junyanz/BicycleGAN |
| contour2im | Smart, Sparse Contours to Represent and Edit Images |
CVPR 2018 | 1712.08232 | website |
分离 | Cross-domain disentanglement networks | Image-to-image translation for cross-domain disentanglement | NIPS 2018 | 1805.09730 |
|
非监督学习
非监督学习- 通用
Note | Model | Paper | Conference | paper link(arXiv) | code link(github) |
---|
| DTN | Unsupervised Cross-Domain Image Generation | ICLR 2017 | 1611.02200 |
yunjey/domain-transfer-network (unofficial) |
| UNIT | Unsupervised image-to-image translation networks | NIPS 2017 | 1703.00848 | mingyuliutw/UNIT |
| DiscoGAN | Learning to Discover Cross-Domain Relations with Generative Adversarial Networks | ICML 2017 | 1703.05192 | SKTBrain/DiscoGAN |
| CycleGAN | Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks | ICCV 2017 | 1703.10593 | junyanz/pytorch-CycleGAN-and-pix2pix |
| DualGAN | DualGAN: Unsupervised Dual Learning for Image-to-Image Translation | ICCV 2017 | 1704.02510 | duxingren14/DualGAN |
| DistanceGAN | One-Sided Unsupervised Domain Mapping | NIPS 2017 | 1706.00826 | sagiebenaim/DistanceGAN |
| Triangle GAN | Triangle Generative Adversarial Networks | NIPS 2017 | 1709.06548 | LiqunChen0606/Triangle-GAN |
特征点导向 | G2-GAN | Geometry Guided Adversarial Facial Expression Synthesis | MM 2018 | 1712.03474 |
|
| CartoonGAN | CartoonGAN: Generative Adversarial Networks for Photo Cartoonization | CVPR 2018 | thecvf | unofficial test, unofficial pytorch |
非对抗 | NAM |
NAM: Non-Adversarial Unsupervised Domain Mapping | ECCV 2018 | 1806.00804 | facebookresearch/nam |
| SCAN | Unsupervised Image-to-Image Translation with Stacked Cycle-Consistent Adversarial Networks | ECCV 2018 | 1807.08536 |
|
空洞卷积,提高形状的变形 | GANimorph | Improved Shape Deformation in Unsupervised Image to Image Translation | ECCV 2018 | 1808.04325 | brownvc/ganimorph |
实例感知 | InstaGAN | Instance-aware image-to-image translation | ICLR 2019 (in review) | openreview |
|
非监督学习- 注意力机制或者模板导向机制
Note | Model | Paper | Conference | paper link(arXiv) | code link(github) |
---|
模板 | ContrastGAN | Generative Semantic Manipulation with Mask-Contrasting GAN | ECCV 2018 | 1708.00315 |
|
注意力机制 | DA-GAN | DA-GAN: Instance-level Image Translation by Deep Attention Generative Adversarial Networks | CVPR 2018 | 1802.06454 |
|
模板/ 注意力 | Attention-GAN | Attention-GAN for Object Transfiguration in Wild Images |
| 1803.06798 |
|
注意力 | Attention guided GAN | Unsupervised Attention-guided Image to Image Translation | NIPS 2018 | 1806.02311 | AlamiMejjati/Unsupervised-Attention-guided-Image-to-Image-Translation |
注意力, 单边 |
| Show, Attend and Translate: Unsupervised Image Translation with Self-Regularization and Attention |
| 1806.06195 |
|
非监督学习-多对多(属性)
Note | Model | Paper | Conference | paper link(arXiv) | code link(github) |
---|
| Conditional CycleGAN | Conditional CycleGAN for Attribute Guided Face Image Generation |
ECCV 2018 | 1705.09966 |
|
| StarGAN | StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation | CVPR 2018 | 1711.09020 | yunjey/StarGAN |
| AttGAN | AttGAN: Facial Attribute Editing by Only Changing What You Want |
| 1711.10678 | LynnHo/AttGAN-Tensorflow |
| ComboGAN | ComboGAN: Unrestrained Scalability for Image Domain Translation | CVPRW 2018 | 1712.06909 | AAnoosheh/ComboGAN |
| AugCGAN (Augmented CycleGAN) | Augmented CycleGAN: Learning Many-to-Many Mappings from Unpaired Data | ICML 2018 | 1802.10151 | aalmah/augmented_cyclegan |
| SG-GAN | Sparsely Grouped Multi-task Generative Adversarial Networks for Facial Attribute Manipulation | MM 2018 | 1805.07509 | zhangqianhui/Sparsely-Grouped-GAN |
| GANimation | GANimation: Anatomically-aware Facial Animation from a Single Image | ECCV 2018 (honorable mention) | 1807.09251 | albertpumarola/GANimation |
非监督学习- 分离(与/或样本导向)
Note | Model | Paper | Conference | paper link(arXiv) | code link(github) |
---|
非分离, 纹理导向 | TextureGAN |
TextureGAN: Controlling Deep Image Synthesis with Texture Patches | CVPR 2018 | 1706.02823 | janesjanes/Pytorch-TextureGAN |
| XGAN | XGAN: Unsupervised Image-to-Image Translation for Many-to-Many Mappings | ICML 2018 | 1711.05139 | dataset |
| ELEGANT | ELEGANT: Exchanging Latent Encodings with GAN for Transferring Multiple Face Attributes | ECCV 2018 | 1803.10562 | Prinsphield/ELEGANT |
| MUNIT | Multimodal Unsupervised Image-to-Image Translation | ECCV 2018 | 1804.04732 | NVlabs/MUNIT |
| cd-GAN (Conditional DualGAN) | Conditional Image-to-Image Translation | CVPR 2018 |
1805.00251 |
|
| EG-UNIT | Exemplar Guided Unsupervised Image-to-Image Translation |
| 1805.11145 |
|
| DRIT | Diverse Image-to-Image Translation via Disentangled Representations | ECCV 2018 | 1808.00948 | HsinYingLee/DRIT |
分分离, 人脸化妆导向 | BeautyGAN | BeautyGAN: Instance-level Facial Makeup Transfer with Deep Generative Adversarial Network | MM 2018 | author |
|
| UFDN | A Unified Feature Disentangler for Multi-Domain Image Translation and Manipulation | NIPS 2018 | 1809.01361 | Alexander-H-Liu/UFDN |
本文选自github
作者:lzhbrian
编译:ronghuaiyang
来源:
https://mp.weixin.qq.com/s/KiIpZb-9vq9bagcLV1aXJQ
*推荐阅读*
PS.本周四(11月15日)晚,TEE首席架构师、TEE AI Lab资深研究员邓文彬将为我们讲解如何在GPU/CPU/移动端高效训练CNN网络,公众号回复“35”即可获取直播详情。左下角阅读原文查看更多直播预告。
