SlideShare a Scribd company logo
Image Inpainting by GAN
Presentation: X37
Date: Mar 22nd, 2020
Place: Tokyo
Overview
• Task overview: What’s the goal of inpainting task?
• Inpainting Task by GAN
- General structures overview
❖ Generator structure
❖ Discriminator structure
• Models
- Partial Convolution
- GLCIC; Globally and Locally Consistent Image Completion
- Contextual Attention
- Gated Convolution
- PEPSI / Diet-PEPSI
Image Inpainting Tasks
Yu, J., Lin, Z., Yang, J., Shen, X., Lu, X., Huang, T. (2018). Free-Form Image Inpainting with Gated Convolutionhttps://arxiv.org/abs/1806.03589
masked	image	
+	mask
Inpainting
model
image
GAN for Image Painting
• Original GAN:
- Generator: I / O: noise / image
- Discriminator: I / O: image / bool (real or fake)
• Conditional GAN:
- Generator: I / O: conditional information / image
- Discriminator: I / O: conditional information + image / bool (real or fake)
DZ G T	/	F T	/	FŶX ( ̂X, Y) DG ̂X
cGAN’s application: Image to Image translation
• Pix2pix generalize the cGAN’s application and its capability:
- synthesizing photos from label maps
- Reconstructing objects from edge maps
- Colorizing images
GAN on Image Inpainting tasks
• Image Inpainting is one of the Image-to-Image translation problems
- Generator: I / O: mask and masked image / inpainted image
- Discriminator: I / O: conditional information + image / bool (real or fake)
T	/	F
full	image	or		
masked	region
DG image
masked	image	
+	mask
• Several study on improvement for GAN
- WGAN, WGAN-GP [Arjovsky+ 2017] [Gulrajani+ 2017]
- Markovian GAN / PatchGAN [Wand 2016]
- Context encoding [Isola+ 2016] [Pathak+ 2016]
- Spectral Normalization / Self-Attention [Miyato+ 2018] [Zhang+ 2019]
• Several study on improvement for each component
- Generator (coarse-to-fine)
- Discriminator (GLCIC; Global and locally consistent image completion, RED)
- Contextual Attention
Image Inpainting for Irregular Holes Using Partial Convolutions
• Summary
- published at Dec 15, 2018
- proposed partial convolution layer
- Inpainting for irregular holes
• Model architecture
UNet-like architecture, replacing all convolutional layers with partial convolutional
layers and using nearest neighbour up-sampling in the decoding stage
• Future work: Loss is complicated, mask update is rule-based
Globally and Locally Consistent Image Completion
• Summary
- Proposed global and local discriminator
- Inpainting for irregular holes
• Model architecture (Auto Encoder based conditional GAN)
- Completion Network (Auto-Encoder based generator)
- Global / Local Discriminator
• Using dilated convolution layers [Yu+ 2016], weighted MSE loss [Pathak+ 2016]
Generative Image Inpainting with Contextual Attention
• Summary
- published at Mar 21, 2018
- proposed contextual attention layer and coarse-to-fine structure
- introduced several GAN’s techniques (GLCIC, WGAN-GP)
• Model architecture: Global and local WGANs
• To measure cosine similarity btw backgrounds and foreground
Free-Form Image Inpainting with Gated Convolution
• Summary
- published at Mar 21, 2018
- proposed the gated convolution for the dynamic mask update
• Model architecture: SN-PatchGAN
- Coarse-to-fine generator with contextual attention
- Spectral-normalized Markovian (Patch) discriminator
PEPSI : Fast Image Inpainting with Parallel Decoding Network
• Summary
- proposed PEPSI: parallel extended-decoder path for semantic inpainting
- Proposed RED(Region Ensemble Discriminator), which adds 1x1 conv on SN-Patch
discriminator
• Model architecture: modified SN-PatchGAN with contextual attention
PEPSI++: Fast and Lightweight Network for Image Inpainting
• Summary
- applied Rate-adaptive dilated conv layers to PEPSI
Appendix
• Dilated convolution
• PatchGAN / MarkovianGAN
• From Pix2pix to cycleGAN
• WGAN / WGAN-GP
Dilated Convolution
• To extract more global features on convolutional layers
PatchGAN / MarkovianGAN
CycleGAN

More Related Content

Similar to 20200322 inpainting

DiscoGAN
DiscoGANDiscoGAN
DiscoGAN
Il Gu Yi
 
From Experimentation to Production: The Future of WebGL
From Experimentation to Production: The Future of WebGLFrom Experimentation to Production: The Future of WebGL
From Experimentation to Production: The Future of WebGL
FITC
 
20191107 deeplearningapproachesfornetworks
20191107 deeplearningapproachesfornetworks20191107 deeplearningapproachesfornetworks
20191107 deeplearningapproachesfornetworks
tm1966
 
187186134 5-geometric-modeling
187186134 5-geometric-modeling187186134 5-geometric-modeling
187186134 5-geometric-modeling
manojg1990
 
187186134 5-geometric-modeling
187186134 5-geometric-modeling187186134 5-geometric-modeling
187186134 5-geometric-modeling
manojg1990
 
5 geometric modeling
5 geometric modeling5 geometric modeling
5 geometric modeling
Ankush Khansole
 
5 geometric-modeling-ppt-university-of-victoria
5 geometric-modeling-ppt-university-of-victoria5 geometric-modeling-ppt-university-of-victoria
5 geometric-modeling-ppt-university-of-victoria
Raghu Gadde
 
Cahall Final Intern Presentation
Cahall Final Intern PresentationCahall Final Intern Presentation
Cahall Final Intern PresentationDaniel Cahall
 
ZJPeng.3DSolderBallReconstruction
ZJPeng.3DSolderBallReconstructionZJPeng.3DSolderBallReconstruction
ZJPeng.3DSolderBallReconstructionZhejian Peng
 
Seeing what a gan cannot generate: paper review
Seeing what a gan cannot generate: paper reviewSeeing what a gan cannot generate: paper review
Seeing what a gan cannot generate: paper review
QuantUniversity
 
2015 crs4-seminar-massive-models-full
2015 crs4-seminar-massive-models-full2015 crs4-seminar-massive-models-full
2015 crs4-seminar-massive-models-full
CRS4 Research Center in Sardinia
 
5_Geometric_Modeling.pdf
5_Geometric_Modeling.pdf5_Geometric_Modeling.pdf
5_Geometric_Modeling.pdf
KeerthanaP37
 
Sticky Notes - a tool for supporting collaborative activities in a 3D virtual...
Sticky Notes - a tool for supporting collaborative activities in a 3D virtual...Sticky Notes - a tool for supporting collaborative activities in a 3D virtual...
Sticky Notes - a tool for supporting collaborative activities in a 3D virtual...
Mikhail Fominykh
 
3D Generalization Lenses (IV 2008)
3D Generalization Lenses (IV 2008)3D Generalization Lenses (IV 2008)
3D Generalization Lenses (IV 2008)Matthias Trapp
 
Decomposing image generation into layout priction and conditional synthesis
Decomposing image generation into layout priction and conditional synthesisDecomposing image generation into layout priction and conditional synthesis
Decomposing image generation into layout priction and conditional synthesis
Naeem Shehzad
 
Authoring of procedural rocks in The Blacksmith realtime short
Authoring of procedural rocks in The Blacksmith realtime shortAuthoring of procedural rocks in The Blacksmith realtime short
Authoring of procedural rocks in The Blacksmith realtime short
Vesselin Efremov
 
Colorspace: Useful For More Than Just Color? - SF Video Tech Meetup - 27 May ...
Colorspace: Useful For More Than Just Color? - SF Video Tech Meetup - 27 May ...Colorspace: Useful For More Than Just Color? - SF Video Tech Meetup - 27 May ...
Colorspace: Useful For More Than Just Color? - SF Video Tech Meetup - 27 May ...
Derek Buitenhuis
 
An Intelligent approach to Pic to Cartoon Conversion using White-box-cartooni...
An Intelligent approach to Pic to Cartoon Conversion using White-box-cartooni...An Intelligent approach to Pic to Cartoon Conversion using White-box-cartooni...
An Intelligent approach to Pic to Cartoon Conversion using White-box-cartooni...
IRJET Journal
 
A Certain Slant of Light - Past, Present and Future Challenges of Global Illu...
A Certain Slant of Light - Past, Present and Future Challenges of Global Illu...A Certain Slant of Light - Past, Present and Future Challenges of Global Illu...
A Certain Slant of Light - Past, Present and Future Challenges of Global Illu...
Electronic Arts / DICE
 
Past, Present and Future Challenges of Global Illumination in Games
Past, Present and Future Challenges of Global Illumination in GamesPast, Present and Future Challenges of Global Illumination in Games
Past, Present and Future Challenges of Global Illumination in Games
Colin Barré-Brisebois
 

Similar to 20200322 inpainting (20)

DiscoGAN
DiscoGANDiscoGAN
DiscoGAN
 
From Experimentation to Production: The Future of WebGL
From Experimentation to Production: The Future of WebGLFrom Experimentation to Production: The Future of WebGL
From Experimentation to Production: The Future of WebGL
 
20191107 deeplearningapproachesfornetworks
20191107 deeplearningapproachesfornetworks20191107 deeplearningapproachesfornetworks
20191107 deeplearningapproachesfornetworks
 
187186134 5-geometric-modeling
187186134 5-geometric-modeling187186134 5-geometric-modeling
187186134 5-geometric-modeling
 
187186134 5-geometric-modeling
187186134 5-geometric-modeling187186134 5-geometric-modeling
187186134 5-geometric-modeling
 
5 geometric modeling
5 geometric modeling5 geometric modeling
5 geometric modeling
 
5 geometric-modeling-ppt-university-of-victoria
5 geometric-modeling-ppt-university-of-victoria5 geometric-modeling-ppt-university-of-victoria
5 geometric-modeling-ppt-university-of-victoria
 
Cahall Final Intern Presentation
Cahall Final Intern PresentationCahall Final Intern Presentation
Cahall Final Intern Presentation
 
ZJPeng.3DSolderBallReconstruction
ZJPeng.3DSolderBallReconstructionZJPeng.3DSolderBallReconstruction
ZJPeng.3DSolderBallReconstruction
 
Seeing what a gan cannot generate: paper review
Seeing what a gan cannot generate: paper reviewSeeing what a gan cannot generate: paper review
Seeing what a gan cannot generate: paper review
 
2015 crs4-seminar-massive-models-full
2015 crs4-seminar-massive-models-full2015 crs4-seminar-massive-models-full
2015 crs4-seminar-massive-models-full
 
5_Geometric_Modeling.pdf
5_Geometric_Modeling.pdf5_Geometric_Modeling.pdf
5_Geometric_Modeling.pdf
 
Sticky Notes - a tool for supporting collaborative activities in a 3D virtual...
Sticky Notes - a tool for supporting collaborative activities in a 3D virtual...Sticky Notes - a tool for supporting collaborative activities in a 3D virtual...
Sticky Notes - a tool for supporting collaborative activities in a 3D virtual...
 
3D Generalization Lenses (IV 2008)
3D Generalization Lenses (IV 2008)3D Generalization Lenses (IV 2008)
3D Generalization Lenses (IV 2008)
 
Decomposing image generation into layout priction and conditional synthesis
Decomposing image generation into layout priction and conditional synthesisDecomposing image generation into layout priction and conditional synthesis
Decomposing image generation into layout priction and conditional synthesis
 
Authoring of procedural rocks in The Blacksmith realtime short
Authoring of procedural rocks in The Blacksmith realtime shortAuthoring of procedural rocks in The Blacksmith realtime short
Authoring of procedural rocks in The Blacksmith realtime short
 
Colorspace: Useful For More Than Just Color? - SF Video Tech Meetup - 27 May ...
Colorspace: Useful For More Than Just Color? - SF Video Tech Meetup - 27 May ...Colorspace: Useful For More Than Just Color? - SF Video Tech Meetup - 27 May ...
Colorspace: Useful For More Than Just Color? - SF Video Tech Meetup - 27 May ...
 
An Intelligent approach to Pic to Cartoon Conversion using White-box-cartooni...
An Intelligent approach to Pic to Cartoon Conversion using White-box-cartooni...An Intelligent approach to Pic to Cartoon Conversion using White-box-cartooni...
An Intelligent approach to Pic to Cartoon Conversion using White-box-cartooni...
 
A Certain Slant of Light - Past, Present and Future Challenges of Global Illu...
A Certain Slant of Light - Past, Present and Future Challenges of Global Illu...A Certain Slant of Light - Past, Present and Future Challenges of Global Illu...
A Certain Slant of Light - Past, Present and Future Challenges of Global Illu...
 
Past, Present and Future Challenges of Global Illumination in Games
Past, Present and Future Challenges of Global Illumination in GamesPast, Present and Future Challenges of Global Illumination in Games
Past, Present and Future Challenges of Global Illumination in Games
 

More from X 37

20200510 37
20200510 3720200510 37
20200510 37
X 37
 
20200209 research note of "superhuman AI for multiplayer poker"
20200209 research note of "superhuman AI for multiplayer poker"20200209 research note of "superhuman AI for multiplayer poker"
20200209 research note of "superhuman AI for multiplayer poker"
X 37
 
20191215 rate distortion theory and VAEs
20191215 rate distortion theory and VAEs20191215 rate distortion theory and VAEs
20191215 rate distortion theory and VAEs
X 37
 
20191027 bread house seminar
20191027 bread house seminar20191027 bread house seminar
20191027 bread house seminar
X 37
 
20190818 Bread Seminar
20190818 Bread Seminar20190818 Bread Seminar
20190818 Bread Seminar
X 37
 
paper repo - pre training for model robustness and uncertainty
paper repo - pre training for model robustness and uncertaintypaper repo - pre training for model robustness and uncertainty
paper repo - pre training for model robustness and uncertainty
X 37
 
Anomaly detection and change detection - sparse structure analysis -
Anomaly detection and change detection - sparse structure analysis -Anomaly detection and change detection - sparse structure analysis -
Anomaly detection and change detection - sparse structure analysis -
X 37
 
Reading papers - survey on Non-Convex Optimization
Reading papers - survey on Non-Convex OptimizationReading papers - survey on Non-Convex Optimization
Reading papers - survey on Non-Convex Optimization
X 37
 
20180520 MLPHS
20180520 MLPHS20180520 MLPHS
20180520 MLPHS
X 37
 

More from X 37 (9)

20200510 37
20200510 3720200510 37
20200510 37
 
20200209 research note of "superhuman AI for multiplayer poker"
20200209 research note of "superhuman AI for multiplayer poker"20200209 research note of "superhuman AI for multiplayer poker"
20200209 research note of "superhuman AI for multiplayer poker"
 
20191215 rate distortion theory and VAEs
20191215 rate distortion theory and VAEs20191215 rate distortion theory and VAEs
20191215 rate distortion theory and VAEs
 
20191027 bread house seminar
20191027 bread house seminar20191027 bread house seminar
20191027 bread house seminar
 
20190818 Bread Seminar
20190818 Bread Seminar20190818 Bread Seminar
20190818 Bread Seminar
 
paper repo - pre training for model robustness and uncertainty
paper repo - pre training for model robustness and uncertaintypaper repo - pre training for model robustness and uncertainty
paper repo - pre training for model robustness and uncertainty
 
Anomaly detection and change detection - sparse structure analysis -
Anomaly detection and change detection - sparse structure analysis -Anomaly detection and change detection - sparse structure analysis -
Anomaly detection and change detection - sparse structure analysis -
 
Reading papers - survey on Non-Convex Optimization
Reading papers - survey on Non-Convex OptimizationReading papers - survey on Non-Convex Optimization
Reading papers - survey on Non-Convex Optimization
 
20180520 MLPHS
20180520 MLPHS20180520 MLPHS
20180520 MLPHS
 

Recently uploaded

PRESENTATION ABOUT PRINCIPLE OF COSMATIC EVALUATION
PRESENTATION ABOUT PRINCIPLE OF COSMATIC EVALUATIONPRESENTATION ABOUT PRINCIPLE OF COSMATIC EVALUATION
PRESENTATION ABOUT PRINCIPLE OF COSMATIC EVALUATION
ChetanK57
 
THE IMPORTANCE OF MARTIAN ATMOSPHERE SAMPLE RETURN.
THE IMPORTANCE OF MARTIAN ATMOSPHERE SAMPLE RETURN.THE IMPORTANCE OF MARTIAN ATMOSPHERE SAMPLE RETURN.
THE IMPORTANCE OF MARTIAN ATMOSPHERE SAMPLE RETURN.
Sérgio Sacani
 
Astronomy Update- Curiosity’s exploration of Mars _ Local Briefs _ leadertele...
Astronomy Update- Curiosity’s exploration of Mars _ Local Briefs _ leadertele...Astronomy Update- Curiosity’s exploration of Mars _ Local Briefs _ leadertele...
Astronomy Update- Curiosity’s exploration of Mars _ Local Briefs _ leadertele...
NathanBaughman3
 
The ASGCT Annual Meeting was packed with exciting progress in the field advan...
The ASGCT Annual Meeting was packed with exciting progress in the field advan...The ASGCT Annual Meeting was packed with exciting progress in the field advan...
The ASGCT Annual Meeting was packed with exciting progress in the field advan...
Health Advances
 
Mammalian Pineal Body Structure and Also Functions
Mammalian Pineal Body Structure and Also FunctionsMammalian Pineal Body Structure and Also Functions
Mammalian Pineal Body Structure and Also Functions
YOGESH DOGRA
 
Nucleic Acid-its structural and functional complexity.
Nucleic Acid-its structural and functional complexity.Nucleic Acid-its structural and functional complexity.
Nucleic Acid-its structural and functional complexity.
Nistarini College, Purulia (W.B) India
 
SCHIZOPHRENIA Disorder/ Brain Disorder.pdf
SCHIZOPHRENIA Disorder/ Brain Disorder.pdfSCHIZOPHRENIA Disorder/ Brain Disorder.pdf
SCHIZOPHRENIA Disorder/ Brain Disorder.pdf
SELF-EXPLANATORY
 
Hemoglobin metabolism_pathophysiology.pptx
Hemoglobin metabolism_pathophysiology.pptxHemoglobin metabolism_pathophysiology.pptx
Hemoglobin metabolism_pathophysiology.pptx
muralinath2
 
What is greenhouse gasses and how many gasses are there to affect the Earth.
What is greenhouse gasses and how many gasses are there to affect the Earth.What is greenhouse gasses and how many gasses are there to affect the Earth.
What is greenhouse gasses and how many gasses are there to affect the Earth.
moosaasad1975
 
role of pramana in research.pptx in science
role of pramana in research.pptx in sciencerole of pramana in research.pptx in science
role of pramana in research.pptx in science
sonaliswain16
 
Hemostasis_importance& clinical significance.pptx
Hemostasis_importance& clinical significance.pptxHemostasis_importance& clinical significance.pptx
Hemostasis_importance& clinical significance.pptx
muralinath2
 
Lab report on liquid viscosity of glycerin
Lab report on liquid viscosity of glycerinLab report on liquid viscosity of glycerin
Lab report on liquid viscosity of glycerin
ossaicprecious19
 
platelets_clotting_biogenesis.clot retractionpptx
platelets_clotting_biogenesis.clot retractionpptxplatelets_clotting_biogenesis.clot retractionpptx
platelets_clotting_biogenesis.clot retractionpptx
muralinath2
 
Richard's entangled aventures in wonderland
Richard's entangled aventures in wonderlandRichard's entangled aventures in wonderland
Richard's entangled aventures in wonderland
Richard Gill
 
Structural Classification Of Protein (SCOP)
Structural Classification Of Protein  (SCOP)Structural Classification Of Protein  (SCOP)
Structural Classification Of Protein (SCOP)
aishnasrivastava
 
ESR_factors_affect-clinic significance-Pathysiology.pptx
ESR_factors_affect-clinic significance-Pathysiology.pptxESR_factors_affect-clinic significance-Pathysiology.pptx
ESR_factors_affect-clinic significance-Pathysiology.pptx
muralinath2
 
general properties of oerganologametal.ppt
general properties of oerganologametal.pptgeneral properties of oerganologametal.ppt
general properties of oerganologametal.ppt
IqrimaNabilatulhusni
 
Unveiling the Energy Potential of Marshmallow Deposits.pdf
Unveiling the Energy Potential of Marshmallow Deposits.pdfUnveiling the Energy Potential of Marshmallow Deposits.pdf
Unveiling the Energy Potential of Marshmallow Deposits.pdf
Erdal Coalmaker
 
RNA INTERFERENCE: UNRAVELING GENETIC SILENCING
RNA INTERFERENCE: UNRAVELING GENETIC SILENCINGRNA INTERFERENCE: UNRAVELING GENETIC SILENCING
RNA INTERFERENCE: UNRAVELING GENETIC SILENCING
AADYARAJPANDEY1
 
GBSN - Microbiology (Lab 4) Culture Media
GBSN - Microbiology (Lab 4) Culture MediaGBSN - Microbiology (Lab 4) Culture Media
GBSN - Microbiology (Lab 4) Culture Media
Areesha Ahmad
 

Recently uploaded (20)

PRESENTATION ABOUT PRINCIPLE OF COSMATIC EVALUATION
PRESENTATION ABOUT PRINCIPLE OF COSMATIC EVALUATIONPRESENTATION ABOUT PRINCIPLE OF COSMATIC EVALUATION
PRESENTATION ABOUT PRINCIPLE OF COSMATIC EVALUATION
 
THE IMPORTANCE OF MARTIAN ATMOSPHERE SAMPLE RETURN.
THE IMPORTANCE OF MARTIAN ATMOSPHERE SAMPLE RETURN.THE IMPORTANCE OF MARTIAN ATMOSPHERE SAMPLE RETURN.
THE IMPORTANCE OF MARTIAN ATMOSPHERE SAMPLE RETURN.
 
Astronomy Update- Curiosity’s exploration of Mars _ Local Briefs _ leadertele...
Astronomy Update- Curiosity’s exploration of Mars _ Local Briefs _ leadertele...Astronomy Update- Curiosity’s exploration of Mars _ Local Briefs _ leadertele...
Astronomy Update- Curiosity’s exploration of Mars _ Local Briefs _ leadertele...
 
The ASGCT Annual Meeting was packed with exciting progress in the field advan...
The ASGCT Annual Meeting was packed with exciting progress in the field advan...The ASGCT Annual Meeting was packed with exciting progress in the field advan...
The ASGCT Annual Meeting was packed with exciting progress in the field advan...
 
Mammalian Pineal Body Structure and Also Functions
Mammalian Pineal Body Structure and Also FunctionsMammalian Pineal Body Structure and Also Functions
Mammalian Pineal Body Structure and Also Functions
 
Nucleic Acid-its structural and functional complexity.
Nucleic Acid-its structural and functional complexity.Nucleic Acid-its structural and functional complexity.
Nucleic Acid-its structural and functional complexity.
 
SCHIZOPHRENIA Disorder/ Brain Disorder.pdf
SCHIZOPHRENIA Disorder/ Brain Disorder.pdfSCHIZOPHRENIA Disorder/ Brain Disorder.pdf
SCHIZOPHRENIA Disorder/ Brain Disorder.pdf
 
Hemoglobin metabolism_pathophysiology.pptx
Hemoglobin metabolism_pathophysiology.pptxHemoglobin metabolism_pathophysiology.pptx
Hemoglobin metabolism_pathophysiology.pptx
 
What is greenhouse gasses and how many gasses are there to affect the Earth.
What is greenhouse gasses and how many gasses are there to affect the Earth.What is greenhouse gasses and how many gasses are there to affect the Earth.
What is greenhouse gasses and how many gasses are there to affect the Earth.
 
role of pramana in research.pptx in science
role of pramana in research.pptx in sciencerole of pramana in research.pptx in science
role of pramana in research.pptx in science
 
Hemostasis_importance& clinical significance.pptx
Hemostasis_importance& clinical significance.pptxHemostasis_importance& clinical significance.pptx
Hemostasis_importance& clinical significance.pptx
 
Lab report on liquid viscosity of glycerin
Lab report on liquid viscosity of glycerinLab report on liquid viscosity of glycerin
Lab report on liquid viscosity of glycerin
 
platelets_clotting_biogenesis.clot retractionpptx
platelets_clotting_biogenesis.clot retractionpptxplatelets_clotting_biogenesis.clot retractionpptx
platelets_clotting_biogenesis.clot retractionpptx
 
Richard's entangled aventures in wonderland
Richard's entangled aventures in wonderlandRichard's entangled aventures in wonderland
Richard's entangled aventures in wonderland
 
Structural Classification Of Protein (SCOP)
Structural Classification Of Protein  (SCOP)Structural Classification Of Protein  (SCOP)
Structural Classification Of Protein (SCOP)
 
ESR_factors_affect-clinic significance-Pathysiology.pptx
ESR_factors_affect-clinic significance-Pathysiology.pptxESR_factors_affect-clinic significance-Pathysiology.pptx
ESR_factors_affect-clinic significance-Pathysiology.pptx
 
general properties of oerganologametal.ppt
general properties of oerganologametal.pptgeneral properties of oerganologametal.ppt
general properties of oerganologametal.ppt
 
Unveiling the Energy Potential of Marshmallow Deposits.pdf
Unveiling the Energy Potential of Marshmallow Deposits.pdfUnveiling the Energy Potential of Marshmallow Deposits.pdf
Unveiling the Energy Potential of Marshmallow Deposits.pdf
 
RNA INTERFERENCE: UNRAVELING GENETIC SILENCING
RNA INTERFERENCE: UNRAVELING GENETIC SILENCINGRNA INTERFERENCE: UNRAVELING GENETIC SILENCING
RNA INTERFERENCE: UNRAVELING GENETIC SILENCING
 
GBSN - Microbiology (Lab 4) Culture Media
GBSN - Microbiology (Lab 4) Culture MediaGBSN - Microbiology (Lab 4) Culture Media
GBSN - Microbiology (Lab 4) Culture Media
 

20200322 inpainting

  • 1. Image Inpainting by GAN Presentation: X37 Date: Mar 22nd, 2020 Place: Tokyo
  • 2. Overview • Task overview: What’s the goal of inpainting task? • Inpainting Task by GAN - General structures overview ❖ Generator structure ❖ Discriminator structure • Models - Partial Convolution - GLCIC; Globally and Locally Consistent Image Completion - Contextual Attention - Gated Convolution - PEPSI / Diet-PEPSI
  • 3. Image Inpainting Tasks Yu, J., Lin, Z., Yang, J., Shen, X., Lu, X., Huang, T. (2018). Free-Form Image Inpainting with Gated Convolutionhttps://arxiv.org/abs/1806.03589 masked image + mask Inpainting model image
  • 4. GAN for Image Painting • Original GAN: - Generator: I / O: noise / image - Discriminator: I / O: image / bool (real or fake) • Conditional GAN: - Generator: I / O: conditional information / image - Discriminator: I / O: conditional information + image / bool (real or fake) DZ G T / F T / FŶX ( ̂X, Y) DG ̂X
  • 5. cGAN’s application: Image to Image translation • Pix2pix generalize the cGAN’s application and its capability: - synthesizing photos from label maps - Reconstructing objects from edge maps - Colorizing images
  • 6. GAN on Image Inpainting tasks • Image Inpainting is one of the Image-to-Image translation problems - Generator: I / O: mask and masked image / inpainted image - Discriminator: I / O: conditional information + image / bool (real or fake) T / F full image or masked region DG image masked image + mask • Several study on improvement for GAN - WGAN, WGAN-GP [Arjovsky+ 2017] [Gulrajani+ 2017] - Markovian GAN / PatchGAN [Wand 2016] - Context encoding [Isola+ 2016] [Pathak+ 2016] - Spectral Normalization / Self-Attention [Miyato+ 2018] [Zhang+ 2019] • Several study on improvement for each component - Generator (coarse-to-fine) - Discriminator (GLCIC; Global and locally consistent image completion, RED) - Contextual Attention
  • 7.
  • 8. Image Inpainting for Irregular Holes Using Partial Convolutions • Summary - published at Dec 15, 2018 - proposed partial convolution layer - Inpainting for irregular holes • Model architecture UNet-like architecture, replacing all convolutional layers with partial convolutional layers and using nearest neighbour up-sampling in the decoding stage • Future work: Loss is complicated, mask update is rule-based
  • 9.
  • 10. Globally and Locally Consistent Image Completion • Summary - Proposed global and local discriminator - Inpainting for irregular holes • Model architecture (Auto Encoder based conditional GAN) - Completion Network (Auto-Encoder based generator) - Global / Local Discriminator • Using dilated convolution layers [Yu+ 2016], weighted MSE loss [Pathak+ 2016]
  • 11.
  • 12. Generative Image Inpainting with Contextual Attention • Summary - published at Mar 21, 2018 - proposed contextual attention layer and coarse-to-fine structure - introduced several GAN’s techniques (GLCIC, WGAN-GP) • Model architecture: Global and local WGANs • To measure cosine similarity btw backgrounds and foreground
  • 13.
  • 14. Free-Form Image Inpainting with Gated Convolution • Summary - published at Mar 21, 2018 - proposed the gated convolution for the dynamic mask update • Model architecture: SN-PatchGAN - Coarse-to-fine generator with contextual attention - Spectral-normalized Markovian (Patch) discriminator
  • 15.
  • 16. PEPSI : Fast Image Inpainting with Parallel Decoding Network • Summary - proposed PEPSI: parallel extended-decoder path for semantic inpainting - Proposed RED(Region Ensemble Discriminator), which adds 1x1 conv on SN-Patch discriminator • Model architecture: modified SN-PatchGAN with contextual attention
  • 17. PEPSI++: Fast and Lightweight Network for Image Inpainting • Summary - applied Rate-adaptive dilated conv layers to PEPSI
  • 18. Appendix • Dilated convolution • PatchGAN / MarkovianGAN • From Pix2pix to cycleGAN • WGAN / WGAN-GP
  • 19. Dilated Convolution • To extract more global features on convolutional layers