SlideShare a Scribd company logo
1 of 10
D S A I Lab.
2021.05.10.
Unpaired Image-to-Image Translation using
Cycle-Consistent Adversarial Networks
Zhu et al., (Berkeley univ.), ICCV 2017
숭실대학교 소프트웨어학과 DSAI Lab. 지승현 발표
1
D S A I Lab.
Index
1. Motivation
2. Method
3. Experiments
4. Summary
2
D S A I Lab.
1. Motivation
– Image translation task
– A task that translate the given image into another one in specific style.
– If there’s enough paired data, this can be done with supervised learning.
– However, getting those is costly.
3
D S A I Lab.
1. Motivation
– “Image generation” approach can be a solution.
– We can assume there’s a pattern in image set 𝒀.
– There’s a underlying relationship between the domains.
– If we can learn
𝐆: 𝐗 → 𝒀, 𝑷 𝒀 𝑿
, we can translate image x into y.
– However, generating images with GAN cannot guarantee
that translated image is similar with original image.
 So this paper proposed ‘Cycle Consistent Translation’
4
D S A I Lab.
1. Motivation
– Proposed method “Cycle consistency” is similar with “Autoencoder”.
– In image translation task, If there’s a translation function
𝑮: 𝑿 → 𝒀 𝑭: 𝒀 → 𝑿
– then cycle consistency loss encourages
𝑭 𝑮 𝒙 ≈ 𝒙 𝑭 𝑮 𝒚 ≈ 𝒚
– In Ideal Autoencoder,
𝑬𝒏𝒄𝒐𝒅𝒆𝒓 𝒙 = 𝒉𝒙 𝑫𝒆𝒄𝒐𝒅𝒆𝒓 𝒉𝒙 = 𝒙
– The key of training process this architecture* is a loss function.
5
* Denoising autoencoder, … etc
D S A I Lab.
2. Method
– Zhu et al. proposed GAN with Cycle consistency loss
– GAN is good idea for training generator.
– However, GAN loss is for generating look-alike-y image, not for assuring y is from x.
– Which means that this can generate random image which meets given condition.
 So Zhu et al. argued that Cycle consistency loss can relieve this problem.
6
D S A I Lab.
2. Method
– Proposed ‘Cycle consistency loss’ :
– This loss helps F(x) to look like that is originated from x.
– Full objective is :
7
D S A I Lab.
3. Experiments
8
D S A I Lab.
3. Experiments
– Some failure cases
– Maybe new method that captures semantic features well is needed.
9
D S A I Lab.
4. Summary
– In image translation, GAN loss cannot assure translated image is
originated from original image. (can generate random image)
– So proposing method ‘cycle consistency loss’ can help model keep
input information.
10

More Related Content

Similar to 220510 지승현 Unpaired Image-to-Image Translation.pptx

ShawnQuinnCSS581FinalProjectReport
ShawnQuinnCSS581FinalProjectReportShawnQuinnCSS581FinalProjectReport
ShawnQuinnCSS581FinalProjectReport
Shawn Quinn
 
deep_stereo_arxiv_2015
deep_stereo_arxiv_2015deep_stereo_arxiv_2015
deep_stereo_arxiv_2015
Ivan Neulander
 
PhotoSketch: Internet Image Montage
PhotoSketch: Internet Image MontagePhotoSketch: Internet Image Montage
PhotoSketch: Internet Image Montage
diTii
 
PhotoSketch: Internet Image Montage
PhotoSketch: Internet Image MontagePhotoSketch: Internet Image Montage
PhotoSketch: Internet Image Montage
diTii
 

Similar to 220510 지승현 Unpaired Image-to-Image Translation.pptx (20)

Learning from Simulated and Unsupervised Images through Adversarial Training....
Learning from Simulated and Unsupervised Images through Adversarial Training....Learning from Simulated and Unsupervised Images through Adversarial Training....
Learning from Simulated and Unsupervised Images through Adversarial Training....
 
Face Recognition Using Gabor features And PCA
Face Recognition Using Gabor features And PCAFace Recognition Using Gabor features And PCA
Face Recognition Using Gabor features And PCA
 
Variants of GANs - Jaejun Yoo
Variants of GANs - Jaejun YooVariants of GANs - Jaejun Yoo
Variants of GANs - Jaejun Yoo
 
IJCER (www.ijceronline.com) International Journal of computational Engineeri...
 IJCER (www.ijceronline.com) International Journal of computational Engineeri... IJCER (www.ijceronline.com) International Journal of computational Engineeri...
IJCER (www.ijceronline.com) International Journal of computational Engineeri...
 
Report face recognition : ArganRecogn
Report face recognition :  ArganRecognReport face recognition :  ArganRecogn
Report face recognition : ArganRecogn
 
Google | Infinite Nature Zero Whitepaper
Google | Infinite Nature Zero WhitepaperGoogle | Infinite Nature Zero Whitepaper
Google | Infinite Nature Zero Whitepaper
 
ShawnQuinnCSS581FinalProjectReport
ShawnQuinnCSS581FinalProjectReportShawnQuinnCSS581FinalProjectReport
ShawnQuinnCSS581FinalProjectReport
 
Rupesh-ibPRIA final
Rupesh-ibPRIA finalRupesh-ibPRIA final
Rupesh-ibPRIA final
 
A comparison between scilab inbuilt module and novel method for image fusion
A comparison between scilab inbuilt module and novel method for image fusionA comparison between scilab inbuilt module and novel method for image fusion
A comparison between scilab inbuilt module and novel method for image fusion
 
Cartoonization of images using machine Learning
Cartoonization of images using machine LearningCartoonization of images using machine Learning
Cartoonization of images using machine Learning
 
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
 
Self-training with Noisy Student improves ImageNet classification
Self-training with Noisy Student improves ImageNet classificationSelf-training with Noisy Student improves ImageNet classification
Self-training with Noisy Student improves ImageNet classification
 
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
 
H0334749
H0334749H0334749
H0334749
 
Generational Adversarial Neural Networks - Essential Reference
Generational Adversarial Neural Networks - Essential ReferenceGenerational Adversarial Neural Networks - Essential Reference
Generational Adversarial Neural Networks - Essential Reference
 
deep_stereo_arxiv_2015
deep_stereo_arxiv_2015deep_stereo_arxiv_2015
deep_stereo_arxiv_2015
 
Image De-Noising Using Deep Neural Network
Image De-Noising Using Deep Neural NetworkImage De-Noising Using Deep Neural Network
Image De-Noising Using Deep Neural Network
 
One shot learning
One shot learningOne shot learning
One shot learning
 
PhotoSketch: Internet Image Montage
PhotoSketch: Internet Image MontagePhotoSketch: Internet Image Montage
PhotoSketch: Internet Image Montage
 
PhotoSketch: Internet Image Montage
PhotoSketch: Internet Image MontagePhotoSketch: Internet Image Montage
PhotoSketch: Internet Image Montage
 

Recently uploaded

Abortion pills in Doha {{ QATAR }} +966572737505) Get Cytotec
Abortion pills in Doha {{ QATAR }} +966572737505) Get CytotecAbortion pills in Doha {{ QATAR }} +966572737505) Get Cytotec
Abortion pills in Doha {{ QATAR }} +966572737505) Get Cytotec
Abortion pills in Riyadh +966572737505 get cytotec
 
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
wsppdmt
 
Client Researchhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh.pptx
Client Researchhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh.pptxClient Researchhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh.pptx
Client Researchhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh.pptx
Stephen266013
 
bams-3rd-case-presentation-scabies-12-05-2020.pptx
bams-3rd-case-presentation-scabies-12-05-2020.pptxbams-3rd-case-presentation-scabies-12-05-2020.pptx
bams-3rd-case-presentation-scabies-12-05-2020.pptx
JocylDuran
 
Audience Researchndfhcvnfgvgbhujhgfv.pptx
Audience Researchndfhcvnfgvgbhujhgfv.pptxAudience Researchndfhcvnfgvgbhujhgfv.pptx
Audience Researchndfhcvnfgvgbhujhgfv.pptx
Stephen266013
 
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
pwgnohujw
 
Abortion pills in Riyadh Saudi Arabia| +966572737505 | Get Cytotec, Unwanted Kit
Abortion pills in Riyadh Saudi Arabia| +966572737505 | Get Cytotec, Unwanted KitAbortion pills in Riyadh Saudi Arabia| +966572737505 | Get Cytotec, Unwanted Kit
Abortion pills in Riyadh Saudi Arabia| +966572737505 | Get Cytotec, Unwanted Kit
Abortion pills in Riyadh +966572737505 get cytotec
 
一比一原版(曼大毕业证书)曼尼托巴大学毕业证成绩单留信学历认证一手价格
一比一原版(曼大毕业证书)曼尼托巴大学毕业证成绩单留信学历认证一手价格一比一原版(曼大毕业证书)曼尼托巴大学毕业证成绩单留信学历认证一手价格
一比一原版(曼大毕业证书)曼尼托巴大学毕业证成绩单留信学历认证一手价格
q6pzkpark
 
Jual obat aborsi Bandung ( 085657271886 ) Cytote pil telat bulan penggugur ka...
Jual obat aborsi Bandung ( 085657271886 ) Cytote pil telat bulan penggugur ka...Jual obat aborsi Bandung ( 085657271886 ) Cytote pil telat bulan penggugur ka...
Jual obat aborsi Bandung ( 085657271886 ) Cytote pil telat bulan penggugur ka...
Klinik kandungan
 

Recently uploaded (20)

Identify Customer Segments to Create Customer Offers for Each Segment - Appli...
Identify Customer Segments to Create Customer Offers for Each Segment - Appli...Identify Customer Segments to Create Customer Offers for Each Segment - Appli...
Identify Customer Segments to Create Customer Offers for Each Segment - Appli...
 
Abortion pills in Doha {{ QATAR }} +966572737505) Get Cytotec
Abortion pills in Doha {{ QATAR }} +966572737505) Get CytotecAbortion pills in Doha {{ QATAR }} +966572737505) Get Cytotec
Abortion pills in Doha {{ QATAR }} +966572737505) Get Cytotec
 
Unsatisfied Bhabhi ℂall Girls Vadodara Book Esha 7427069034 Top Class ℂall Gi...
Unsatisfied Bhabhi ℂall Girls Vadodara Book Esha 7427069034 Top Class ℂall Gi...Unsatisfied Bhabhi ℂall Girls Vadodara Book Esha 7427069034 Top Class ℂall Gi...
Unsatisfied Bhabhi ℂall Girls Vadodara Book Esha 7427069034 Top Class ℂall Gi...
 
Bios of leading Astrologers & Researchers
Bios of leading Astrologers & ResearchersBios of leading Astrologers & Researchers
Bios of leading Astrologers & Researchers
 
Harnessing the Power of GenAI for BI and Reporting.pptx
Harnessing the Power of GenAI for BI and Reporting.pptxHarnessing the Power of GenAI for BI and Reporting.pptx
Harnessing the Power of GenAI for BI and Reporting.pptx
 
社内勉強会資料_Object Recognition as Next Token Prediction
社内勉強会資料_Object Recognition as Next Token Prediction社内勉強会資料_Object Recognition as Next Token Prediction
社内勉強会資料_Object Recognition as Next Token Prediction
 
SCI8-Q4-MOD11.pdfwrwujrrjfaajerjrajrrarj
SCI8-Q4-MOD11.pdfwrwujrrjfaajerjrajrrarjSCI8-Q4-MOD11.pdfwrwujrrjfaajerjrajrrarj
SCI8-Q4-MOD11.pdfwrwujrrjfaajerjrajrrarj
 
SAC 25 Final National, Regional & Local Angel Group Investing Insights 2024 0...
SAC 25 Final National, Regional & Local Angel Group Investing Insights 2024 0...SAC 25 Final National, Regional & Local Angel Group Investing Insights 2024 0...
SAC 25 Final National, Regional & Local Angel Group Investing Insights 2024 0...
 
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
如何办理英国诺森比亚大学毕业证(NU毕业证书)成绩单原件一模一样
 
Client Researchhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh.pptx
Client Researchhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh.pptxClient Researchhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh.pptx
Client Researchhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh.pptx
 
bams-3rd-case-presentation-scabies-12-05-2020.pptx
bams-3rd-case-presentation-scabies-12-05-2020.pptxbams-3rd-case-presentation-scabies-12-05-2020.pptx
bams-3rd-case-presentation-scabies-12-05-2020.pptx
 
Audience Researchndfhcvnfgvgbhujhgfv.pptx
Audience Researchndfhcvnfgvgbhujhgfv.pptxAudience Researchndfhcvnfgvgbhujhgfv.pptx
Audience Researchndfhcvnfgvgbhujhgfv.pptx
 
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
 
Jual Obat Aborsi Bandung (Asli No.1) Wa 082134680322 Klinik Obat Penggugur Ka...
Jual Obat Aborsi Bandung (Asli No.1) Wa 082134680322 Klinik Obat Penggugur Ka...Jual Obat Aborsi Bandung (Asli No.1) Wa 082134680322 Klinik Obat Penggugur Ka...
Jual Obat Aborsi Bandung (Asli No.1) Wa 082134680322 Klinik Obat Penggugur Ka...
 
Abortion pills in Riyadh Saudi Arabia| +966572737505 | Get Cytotec, Unwanted Kit
Abortion pills in Riyadh Saudi Arabia| +966572737505 | Get Cytotec, Unwanted KitAbortion pills in Riyadh Saudi Arabia| +966572737505 | Get Cytotec, Unwanted Kit
Abortion pills in Riyadh Saudi Arabia| +966572737505 | Get Cytotec, Unwanted Kit
 
一比一原版(曼大毕业证书)曼尼托巴大学毕业证成绩单留信学历认证一手价格
一比一原版(曼大毕业证书)曼尼托巴大学毕业证成绩单留信学历认证一手价格一比一原版(曼大毕业证书)曼尼托巴大学毕业证成绩单留信学历认证一手价格
一比一原版(曼大毕业证书)曼尼托巴大学毕业证成绩单留信学历认证一手价格
 
DBMS UNIT 5 46 CONTAINS NOTES FOR THE STUDENTS
DBMS UNIT 5 46 CONTAINS NOTES FOR THE STUDENTSDBMS UNIT 5 46 CONTAINS NOTES FOR THE STUDENTS
DBMS UNIT 5 46 CONTAINS NOTES FOR THE STUDENTS
 
Jual obat aborsi Bandung ( 085657271886 ) Cytote pil telat bulan penggugur ka...
Jual obat aborsi Bandung ( 085657271886 ) Cytote pil telat bulan penggugur ka...Jual obat aborsi Bandung ( 085657271886 ) Cytote pil telat bulan penggugur ka...
Jual obat aborsi Bandung ( 085657271886 ) Cytote pil telat bulan penggugur ka...
 
jll-asia-pacific-capital-tracker-1q24.pdf
jll-asia-pacific-capital-tracker-1q24.pdfjll-asia-pacific-capital-tracker-1q24.pdf
jll-asia-pacific-capital-tracker-1q24.pdf
 
Digital Transformation Playbook by Graham Ware
Digital Transformation Playbook by Graham WareDigital Transformation Playbook by Graham Ware
Digital Transformation Playbook by Graham Ware
 

220510 지승현 Unpaired Image-to-Image Translation.pptx

  • 1. D S A I Lab. 2021.05.10. Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks Zhu et al., (Berkeley univ.), ICCV 2017 숭실대학교 소프트웨어학과 DSAI Lab. 지승현 발표 1
  • 2. D S A I Lab. Index 1. Motivation 2. Method 3. Experiments 4. Summary 2
  • 3. D S A I Lab. 1. Motivation – Image translation task – A task that translate the given image into another one in specific style. – If there’s enough paired data, this can be done with supervised learning. – However, getting those is costly. 3
  • 4. D S A I Lab. 1. Motivation – “Image generation” approach can be a solution. – We can assume there’s a pattern in image set 𝒀. – There’s a underlying relationship between the domains. – If we can learn 𝐆: 𝐗 → 𝒀, 𝑷 𝒀 𝑿 , we can translate image x into y. – However, generating images with GAN cannot guarantee that translated image is similar with original image.  So this paper proposed ‘Cycle Consistent Translation’ 4
  • 5. D S A I Lab. 1. Motivation – Proposed method “Cycle consistency” is similar with “Autoencoder”. – In image translation task, If there’s a translation function 𝑮: 𝑿 → 𝒀 𝑭: 𝒀 → 𝑿 – then cycle consistency loss encourages 𝑭 𝑮 𝒙 ≈ 𝒙 𝑭 𝑮 𝒚 ≈ 𝒚 – In Ideal Autoencoder, 𝑬𝒏𝒄𝒐𝒅𝒆𝒓 𝒙 = 𝒉𝒙 𝑫𝒆𝒄𝒐𝒅𝒆𝒓 𝒉𝒙 = 𝒙 – The key of training process this architecture* is a loss function. 5 * Denoising autoencoder, … etc
  • 6. D S A I Lab. 2. Method – Zhu et al. proposed GAN with Cycle consistency loss – GAN is good idea for training generator. – However, GAN loss is for generating look-alike-y image, not for assuring y is from x. – Which means that this can generate random image which meets given condition.  So Zhu et al. argued that Cycle consistency loss can relieve this problem. 6
  • 7. D S A I Lab. 2. Method – Proposed ‘Cycle consistency loss’ : – This loss helps F(x) to look like that is originated from x. – Full objective is : 7
  • 8. D S A I Lab. 3. Experiments 8
  • 9. D S A I Lab. 3. Experiments – Some failure cases – Maybe new method that captures semantic features well is needed. 9
  • 10. D S A I Lab. 4. Summary – In image translation, GAN loss cannot assure translated image is originated from original image. (can generate random image) – So proposing method ‘cycle consistency loss’ can help model keep input information. 10