SlideShare a Scribd company logo
CC6 C 2B6 2 C D6 6 6 C 6A2C B A 2 6
.6BC A2C C96 -A6B6 6 0 3 6 BC AC B
I
1-.
•
• ) [ V
• A O LIR +1 ]LIDB E LC
• ( A O B E LC
• = +
. G M
8 7 )9
P . G M
81 2 (9
P .
80 , 9
@
- - - - - - -
- - - - - -
• , 3 3
•
A
Input
Ours
Ground
truth
Mix Raindrop Blur Noise JPEG
SE R T UA PXC I MKLNE O
-
•
• 4
- -
- -
D
• B
• 5
•
+
,,K QHN
• . 6
•
VRP ILM
2 0
2 0 + C 2 C 0
, 219 8
C C
NO P
, ILE A K D I EI DI E
I ED E C IE I ED
• D DE 0 D C D 0 D 1 A M A A I D
• 1E EA 2D K IM ,
• 9 E A 7
• 5
2 8 810 1 , 1
2 8 810 881 82 1 , 1
9:
I
(
op1
op2
op3
op4
op1
op2
op3
op4
…
Input
op1
op2
op3
op4
)1 2
D A F
• 3 03
• 3
• 0
3 1 3 1
-
• 2 ( ) 6
• 2 ( ) 12 3×3
• 3 - )2 3×3
• ( 1 ) (
-
FeatureExtraction
Block
Operation-wise
AttentionLayer
Operation-wise
AttentionLayer
Reconstruction
Layer
Operation-wise
AttentionLayer
•
• 1
1
3
3
AC
+ 4 18C 7E 8 8 F 8
CC 01
. ,E 2AE88H8 7 8 F C 01
! ∈ ℝ$
• 5%&'( ∈ ℝ)×+×$ - ! ∈ ℝ$
• 1
!" = [%&"
'
, … , %&"
*
]
•
,- ∈ ℝ0×2, ,3 ∈ ℝ|*|×05" 6" = ,78(,'6")
%&"
;
=
exp(5"(6"))
∑ exp(5"(6"))
=
exp(&"
;
)
∑;@'
|*|
exp(&"
;
)
→ B C 6 1
! "#
• $%&'(')*% +,-. 1×1
• $%&'(')*% +,-. 3×3
• $%&'(')*% +,-. 5×5
• $%&'(')*% +,-. 7×7
• $%&'(')*% 45*'6%4 +,-. 3×3 (( = 2)
• $%&'(')*% 45*'6%4 +,-. 5×5 (( = 2)
• $%&'(')*% 45*'6%4 +,-. 7×7 (( = 2)
• '.%(';% &,,*5-; 3×3
!" #"
$×$
. 1×1
ℱ(: 1×1 *+,-
• - = 1
• 4 9 =
• =
Feature maps
!×#×$
%&'(
)&
*+,-(
Feature maps
!×#×$
%&
Feature maps
!×#×$
%&-,'(
…
OperationLayerAttentionLayer
…
Concat
1×1conv.
/&
OperationLayer
Concat
1×1conv.
*+,-,
• - -- - 2 0 -
Feature maps
!×#×$
%&'(
)&
*+,-(
Feature maps
!×#×$
%&
Feature maps
!×#×$
%&-,'(
…
OperationLayerAttentionLayer
…
Concat
1×1conv.
/&
OperationLayer
Concat
1×1conv.
*+,-,
• 2
• 1 3 1 23 1 3 ,
• 2 +
• 1 3 1 23 1 3 ,
• 2
• 1 1 1 3 1 23 1 3 , 3 312
,
• D : 8 : :
• 63×63 C I: PYV R[V K V ]
• : + C 8 : : : + C 8 :
• 2 b
• a 4 5 9 ]
,
, :
- :9
4 5 3 , 8 : : 9::C : 8: : : . ,201
02.1 22 D IK
16 8 76 3 , 54 4 6
EP PE 6I D 6MDE E :ESE E
6E IC 8:79 ::36 8:79 ::36 8:79 ::36
1L077 ) ) ) ) ) )- (
95 9EP M E , ( () )) ) )- )
4 L E ETMLD RPPI L DELMIPE . 9EPIDR E LIL M DEEN CLL M IK E DELMIPIL 3222 LP C IMLP ML 3K E 8 MCEPPIL . ( ))
4 R E 0 IL MM C IL M IK E EP M IML BT DEEN EIL M CEKEL E LIL 3L 0 89 ,
• v tnr 077edf
• v uo 8 E ILED 077cighm]aYZp VZs bl [
u DIP M IML l
1 4
- 4 2-
2 42 4
PS IG 5 CD 5D R G PSV
JL O
• 1(3)( 0) 5
• 33 A D 5
• ,5ECC 5 C ,5ECC 5 E 1 , 7 A CC 5 D 57DC
( CD D C 2 CD
0E D 7D
] V [cedf kgn 73 7 LMIK ob a
• h mi P I K LMIK M I Yl aYR
182785 6 1 76 .41 01 1 376
/6 KI K I M IMM L K
(. ) ) ) , . ( ( ( . )-
,) . ,- % % , .
% ) ), - %
(, (. - ( , ( . , ) ( )
,( ) ,( - ) ) , , . (
% % %
M C K IP M I CIKL KLI M LC LIA MK MO
(( , . ) ( )( . ) )) - ( () ) -
- ( ) , ,- - - ( % . , , -
% - , % ) - %
) - ( (. ( ( ) , )) , () , )( ) ( (( - ). -
. - ) - - - . . , , ( , . ( ) ( ( , .
% % %
850
,
89
73 7 LMIK
P I K LMIK M I
89
73 7 LMIK
P I K LMIK M I
4 MCI
2 9 M 0K AM MII C AIK K LMIK M I K AIK M K 1 0867 -
Cat
Cat
Cat
Sheep
Person
Person
Person
Person
Boat
Dog
Dog
Person
Cow
Cow
Dog
Dog
Dog
Dog
Person
Person
Person
Person
Person
Dog
Bottle
Bottle
Clean
Distorted
RL-Restore [1]
Ours
Cat
Person
Person
Person
Person
Person
Person
Person
Person
Cat
TVmonitor
Person
Person
Person
Person
Horse
Horse Horse
HorseCar
I JE A G
• .2 2 2A5 0 1 IRQ
• 128×128 5 I
• 2 A5 2 385A 5A A5 2 385A
• ( , ) I P
)
0 1 . 2 5 2 5 C5 5 5 2 C5 2 C5 A2 2 5 D : 2 5 C2 2 A 5 2 5 ,.
52 A 5
• S - . 2
/
• - . 2 >9 - - 9 D
• - - - -
• a - G
A
• RE N g
Ga e i J
• 1 h RE
N
• 33 3
33 3 P d J
1 E B P
G
: 9 58
2 199A () (- . -.
4429 TA) () . .(
3 103A ( - .(+
9
(
4 351 04 89 48
: 9 58
2 199A ( ) -- ,
8 N9 TA+ ( - -,-
9
. 4. 84 1 9168 .5 .9 42. 9168 .5
SK IM FKSTPRTKP h] bifkald [c eg
A) 6 C I 4 F T P CRF C CST C F M KDM SPM TKP PR E DCS F KNCI F PKSK I 5333 RC SCETKP S P 5NCI :RPE SSK I , . )+ -Z)+ -
A 8 IC NC 3 MPKTK I T PT TKCM P STC FCRF EP VPM TKP CM C TP EPF RS PR KNCI R STPRCTKP DY VPM TKP CRY S CRE 5 5187 -
A+ CK 8 N T 0 RSKST T N NPRY T PRL PR KNCI R STPRCTKP 5 511 ,
• 2CC C a
• t gec gecVy ] R
• C V s
• kbfapKy 0 1 Y I- V r R
• u 2CC C a R - 2 3 Y]K
ldm n hio c P [ VN L[
0 1 C 2 2 C 8 2 C 2 28 C 2C 3 C 2 8 .,-
•
• - -
• e r p
• D V I
• - 43 l I n
• - d 4 - ? e o m
•
• b g h d 4 a - -
i
• - -

More Related Content

What's hot

Математическая модель системы оценки эффективности реализации программ развит...
Математическая модель системы оценки эффективности реализации программ развит...Математическая модель системы оценки эффективности реализации программ развит...
Математическая модель системы оценки эффективности реализации программ развит...
Anatoly Simkin
 
【修士論文紹介】ソーシャルメディアからの作用を考慮した金融市場の観測・予測モデルの提案
【修士論文紹介】ソーシャルメディアからの作用を考慮した金融市場の観測・予測モデルの提案【修士論文紹介】ソーシャルメディアからの作用を考慮した金融市場の観測・予測モデルの提案
【修士論文紹介】ソーシャルメディアからの作用を考慮した金融市場の観測・予測モデルの提案
Koichiro tamura
 
中学生を対象とした 「数学と人間の活動」の授業開発 ー「石取りゲーム」を教材としてー
中学生を対象とした「数学と人間の活動」の授業開発ー「石取りゲーム」を教材としてー中学生を対象とした「数学と人間の活動」の授業開発ー「石取りゲーム」を教材としてー
中学生を対象とした 「数学と人間の活動」の授業開発 ー「石取りゲーム」を教材としてー
Tomomi Furubayashi
 
Spectrum sensing based on goodness of fit test with unilateral alternative hy...
Spectrum sensing based on goodness of fit test with unilateral alternative hy...Spectrum sensing based on goodness of fit test with unilateral alternative hy...
Spectrum sensing based on goodness of fit test with unilateral alternative hy...
Devanshi Piprottar
 
異常検知とGAN: Efficient GAN
異常検知とGAN: Efficient GAN異常検知とGAN: Efficient GAN
異常検知とGAN: Efficient GAN
Koichiro tamura
 
Encrypting sensitive data for puppet
Encrypting sensitive data for puppetEncrypting sensitive data for puppet
Encrypting sensitive data for puppet
sihil
 
[DL Hacks]Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternati...
[DL Hacks]Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternati...[DL Hacks]Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternati...
[DL Hacks]Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternati...
Deep Learning JP
 
ใบงานที่ 4
ใบงานที่ 4ใบงานที่ 4
ใบงานที่ 4
Kanjana Kammek
 
Stalin and soviet industrialisation vox, cepr policy portal
Stalin and soviet industrialisation   vox, cepr policy portalStalin and soviet industrialisation   vox, cepr policy portal
Stalin and soviet industrialisation vox, cepr policy portal
tapanma
 
Determinacion minimos cuadrados
Determinacion minimos cuadradosDeterminacion minimos cuadrados
Determinacion minimos cuadrados
Spier R. B. Condori Cruz
 
창의력 향상을 위한 마인드맵 활용하기
창의력 향상을 위한 마인드맵 활용하기창의력 향상을 위한 마인드맵 활용하기
창의력 향상을 위한 마인드맵 활용하기
Jinho Jung
 
Altivar
AltivarAltivar
[DL Hacks]Pruning Convolutional Neural Networks for Resource Efficient Inference
[DL Hacks]Pruning Convolutional Neural Networks for Resource Efficient Inference[DL Hacks]Pruning Convolutional Neural Networks for Resource Efficient Inference
[DL Hacks]Pruning Convolutional Neural Networks for Resource Efficient Inference
Deep Learning JP
 
Spark MLlib ML Pipelines の概要 及びpysparkからの扱い方
Spark MLlib ML Pipelines の概要 及びpysparkからの扱い方Spark MLlib ML Pipelines の概要 及びpysparkからの扱い方
Spark MLlib ML Pipelines の概要 及びpysparkからの扱い方
Takeshi Mikami
 
Ejercicio 8
Ejercicio 8Ejercicio 8
Ejercicio 8
hermesortiz1508
 
залікове інтеграли
залікове інтегрализалікове інтеграли
залікове інтеграли
Phaust94
 
アーリース情報技術株式会社 会社案内 (2019/02/13)
アーリース情報技術株式会社 会社案内 (2019/02/13)アーリース情報技術株式会社 会社案内 (2019/02/13)
アーリース情報技術株式会社 会社案内 (2019/02/13)
Takeshi Mikami
 
eGovinnovation - Elenea Mugellini
eGovinnovation - Elenea MugellinieGovinnovation - Elenea Mugellini
eGovinnovation - Elenea MugelliniTechnoArk
 

What's hot (20)

Математическая модель системы оценки эффективности реализации программ развит...
Математическая модель системы оценки эффективности реализации программ развит...Математическая модель системы оценки эффективности реализации программ развит...
Математическая модель системы оценки эффективности реализации программ развит...
 
【修士論文紹介】ソーシャルメディアからの作用を考慮した金融市場の観測・予測モデルの提案
【修士論文紹介】ソーシャルメディアからの作用を考慮した金融市場の観測・予測モデルの提案【修士論文紹介】ソーシャルメディアからの作用を考慮した金融市場の観測・予測モデルの提案
【修士論文紹介】ソーシャルメディアからの作用を考慮した金融市場の観測・予測モデルの提案
 
Arran Map
Arran MapArran Map
Arran Map
 
中学生を対象とした 「数学と人間の活動」の授業開発 ー「石取りゲーム」を教材としてー
中学生を対象とした「数学と人間の活動」の授業開発ー「石取りゲーム」を教材としてー中学生を対象とした「数学と人間の活動」の授業開発ー「石取りゲーム」を教材としてー
中学生を対象とした 「数学と人間の活動」の授業開発 ー「石取りゲーム」を教材としてー
 
Spectrum sensing based on goodness of fit test with unilateral alternative hy...
Spectrum sensing based on goodness of fit test with unilateral alternative hy...Spectrum sensing based on goodness of fit test with unilateral alternative hy...
Spectrum sensing based on goodness of fit test with unilateral alternative hy...
 
異常検知とGAN: Efficient GAN
異常検知とGAN: Efficient GAN異常検知とGAN: Efficient GAN
異常検知とGAN: Efficient GAN
 
Encrypting sensitive data for puppet
Encrypting sensitive data for puppetEncrypting sensitive data for puppet
Encrypting sensitive data for puppet
 
[DL Hacks]Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternati...
[DL Hacks]Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternati...[DL Hacks]Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternati...
[DL Hacks]Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternati...
 
ใบงานที่ 4
ใบงานที่ 4ใบงานที่ 4
ใบงานที่ 4
 
Stalin and soviet industrialisation vox, cepr policy portal
Stalin and soviet industrialisation   vox, cepr policy portalStalin and soviet industrialisation   vox, cepr policy portal
Stalin and soviet industrialisation vox, cepr policy portal
 
Determinacion minimos cuadrados
Determinacion minimos cuadradosDeterminacion minimos cuadrados
Determinacion minimos cuadrados
 
창의력 향상을 위한 마인드맵 활용하기
창의력 향상을 위한 마인드맵 활용하기창의력 향상을 위한 마인드맵 활용하기
창의력 향상을 위한 마인드맵 활용하기
 
Altivar
AltivarAltivar
Altivar
 
[DL Hacks]Pruning Convolutional Neural Networks for Resource Efficient Inference
[DL Hacks]Pruning Convolutional Neural Networks for Resource Efficient Inference[DL Hacks]Pruning Convolutional Neural Networks for Resource Efficient Inference
[DL Hacks]Pruning Convolutional Neural Networks for Resource Efficient Inference
 
Spark MLlib ML Pipelines の概要 及びpysparkからの扱い方
Spark MLlib ML Pipelines の概要 及びpysparkからの扱い方Spark MLlib ML Pipelines の概要 及びpysparkからの扱い方
Spark MLlib ML Pipelines の概要 及びpysparkからの扱い方
 
Ejercicio 8
Ejercicio 8Ejercicio 8
Ejercicio 8
 
залікове інтеграли
залікове інтегрализалікове інтеграли
залікове інтеграли
 
Noveno1 p hugo
Noveno1 p hugoNoveno1 p hugo
Noveno1 p hugo
 
アーリース情報技術株式会社 会社案内 (2019/02/13)
アーリース情報技術株式会社 会社案内 (2019/02/13)アーリース情報技術株式会社 会社案内 (2019/02/13)
アーリース情報技術株式会社 会社案内 (2019/02/13)
 
eGovinnovation - Elenea Mugellini
eGovinnovation - Elenea MugellinieGovinnovation - Elenea Mugellini
eGovinnovation - Elenea Mugellini
 

Similar to Attention-Based Adaptive Selection of Operations for Image Restoration in the Presence of Unknown Combined Distortions [CVPR 2019 読み会]

katagaitai CTF workshop #10 AESに対する相関電力解析
katagaitai CTF workshop #10 AESに対する相関電力解析katagaitai CTF workshop #10 AESに対する相関電力解析
katagaitai CTF workshop #10 AESに対する相関電力解析
trmr
 
技術とデザインの最適な関係; 技術の意味を与えるデザイン
技術とデザインの最適な関係; 技術の意味を与えるデザイン技術とデザインの最適な関係; 技術の意味を与えるデザイン
技術とデザインの最適な関係; 技術の意味を与えるデザイン
Tohru Yoshioka-Kobayashi
 
Functional Gradient Boosting based on Residual Network Perception
Functional Gradient Boosting based on Residual Network PerceptionFunctional Gradient Boosting based on Residual Network Perception
Functional Gradient Boosting based on Residual Network Perception
Atsushi Nitanda
 
A research paper introduction of Universal transformers
A research paper introduction of Universal transformersA research paper introduction of Universal transformers
A research paper introduction of Universal transformers
Keigo Kawamura
 
Prelude to halide_public
Prelude to halide_publicPrelude to halide_public
Prelude to halide_public
Fixstars Corporation
 
Distributed Denial of Service Attack Prevention at Source Machines
Distributed Denial of Service Attack Prevention at Source MachinesDistributed Denial of Service Attack Prevention at Source Machines
Distributed Denial of Service Attack Prevention at Source Machines
Shinagawa Laboratory, The University of Tokyo
 
【CVPR 2019】Learning Cross Modal Embeddings with Adversarial Networks for Cook...
【CVPR 2019】Learning Cross Modal Embeddings with Adversarial Networks for Cook...【CVPR 2019】Learning Cross Modal Embeddings with Adversarial Networks for Cook...
【CVPR 2019】Learning Cross Modal Embeddings with Adversarial Networks for Cook...
cvpaper. challenge
 
Наибольшая общая мера: 2500 лет
Наибольшая общая мера: 2500 летНаибольшая общая мера: 2500 лет
Наибольшая общая мера: 2500 летsixtyone
 
JTF2018_B30_k8s_operator_nobusue
JTF2018_B30_k8s_operator_nobusueJTF2018_B30_k8s_operator_nobusue
JTF2018_B30_k8s_operator_nobusue
Nobuhiro Sue
 
vcdで日本語(3) long format が旧世界とのGateway
vcdで日本語(3) long format が旧世界とのGatewayvcdで日本語(3) long format が旧世界とのGateway
vcdで日本語(3) long format が旧世界とのGateway
Tsuda University Institute for Mathematics and Computer Science
 
【ECCV 2018】CornerNet: Detecting Objects as Paired Keypoints
【ECCV 2018】CornerNet: Detecting Objects as Paired Keypoints【ECCV 2018】CornerNet: Detecting Objects as Paired Keypoints
【ECCV 2018】CornerNet: Detecting Objects as Paired Keypoints
cvpaper. challenge
 
深層学習による非滑らかな関数の推定
深層学習による非滑らかな関数の推定深層学習による非滑らかな関数の推定
深層学習による非滑らかな関数の推定
Masaaki Imaizumi
 
Cameroun - Repertoire des projets prioritaires à besoins de financement
Cameroun - Repertoire des projets prioritaires à besoins de financementCameroun - Repertoire des projets prioritaires à besoins de financement
Cameroun - Repertoire des projets prioritaires à besoins de financementinvestincameroon
 
kintone on EKS ― EKS で実現するインフラ自動構築パイプライン
kintone on EKS ― EKS で実現するインフラ自動構築パイプライン kintone on EKS ― EKS で実現するインフラ自動構築パイプライン
kintone on EKS ― EKS で実現するインフラ自動構築パイプライン
Yusuke Nojima
 
第2回 基本演算,データ型の基礎,ベクトルの操作方法
第2回 基本演算,データ型の基礎,ベクトルの操作方法第2回 基本演算,データ型の基礎,ベクトルの操作方法
第2回 基本演算,データ型の基礎,ベクトルの操作方法
Wataru Shito
 
إجراءات قائمة الشحن
إجراءات قائمة الشحنإجراءات قائمة الشحن
إجراءات قائمة الشحن
Mohamed Hendawy
 
音響信号に対する異常音検知技術と応用
音響信号に対する異常音検知技術と応用音響信号に対する異常音検知技術と応用
音響信号に対する異常音検知技術と応用
Yuma Koizumi
 
機械学習と自動微分
機械学習と自動微分機械学習と自動微分
機械学習と自動微分
Ichigaku Takigawa
 

Similar to Attention-Based Adaptive Selection of Operations for Image Restoration in the Presence of Unknown Combined Distortions [CVPR 2019 読み会] (20)

katagaitai CTF workshop #10 AESに対する相関電力解析
katagaitai CTF workshop #10 AESに対する相関電力解析katagaitai CTF workshop #10 AESに対する相関電力解析
katagaitai CTF workshop #10 AESに対する相関電力解析
 
技術とデザインの最適な関係; 技術の意味を与えるデザイン
技術とデザインの最適な関係; 技術の意味を与えるデザイン技術とデザインの最適な関係; 技術の意味を与えるデザイン
技術とデザインの最適な関係; 技術の意味を与えるデザイン
 
Functional Gradient Boosting based on Residual Network Perception
Functional Gradient Boosting based on Residual Network PerceptionFunctional Gradient Boosting based on Residual Network Perception
Functional Gradient Boosting based on Residual Network Perception
 
A research paper introduction of Universal transformers
A research paper introduction of Universal transformersA research paper introduction of Universal transformers
A research paper introduction of Universal transformers
 
Prelude to halide_public
Prelude to halide_publicPrelude to halide_public
Prelude to halide_public
 
Distributed Denial of Service Attack Prevention at Source Machines
Distributed Denial of Service Attack Prevention at Source MachinesDistributed Denial of Service Attack Prevention at Source Machines
Distributed Denial of Service Attack Prevention at Source Machines
 
【CVPR 2019】Learning Cross Modal Embeddings with Adversarial Networks for Cook...
【CVPR 2019】Learning Cross Modal Embeddings with Adversarial Networks for Cook...【CVPR 2019】Learning Cross Modal Embeddings with Adversarial Networks for Cook...
【CVPR 2019】Learning Cross Modal Embeddings with Adversarial Networks for Cook...
 
Наибольшая общая мера: 2500 лет
Наибольшая общая мера: 2500 летНаибольшая общая мера: 2500 лет
Наибольшая общая мера: 2500 лет
 
JTF2018_B30_k8s_operator_nobusue
JTF2018_B30_k8s_operator_nobusueJTF2018_B30_k8s_operator_nobusue
JTF2018_B30_k8s_operator_nobusue
 
vcdで日本語(3) long format が旧世界とのGateway
vcdで日本語(3) long format が旧世界とのGatewayvcdで日本語(3) long format が旧世界とのGateway
vcdで日本語(3) long format が旧世界とのGateway
 
【ECCV 2018】CornerNet: Detecting Objects as Paired Keypoints
【ECCV 2018】CornerNet: Detecting Objects as Paired Keypoints【ECCV 2018】CornerNet: Detecting Objects as Paired Keypoints
【ECCV 2018】CornerNet: Detecting Objects as Paired Keypoints
 
深層学習による非滑らかな関数の推定
深層学習による非滑らかな関数の推定深層学習による非滑らかな関数の推定
深層学習による非滑らかな関数の推定
 
Cameroun - Repertoire des projets prioritaires à besoins de financement
Cameroun - Repertoire des projets prioritaires à besoins de financementCameroun - Repertoire des projets prioritaires à besoins de financement
Cameroun - Repertoire des projets prioritaires à besoins de financement
 
kintone on EKS ― EKS で実現するインフラ自動構築パイプライン
kintone on EKS ― EKS で実現するインフラ自動構築パイプライン kintone on EKS ― EKS で実現するインフラ自動構築パイプライン
kintone on EKS ― EKS で実現するインフラ自動構築パイプライン
 
第2回 基本演算,データ型の基礎,ベクトルの操作方法
第2回 基本演算,データ型の基礎,ベクトルの操作方法第2回 基本演算,データ型の基礎,ベクトルの操作方法
第2回 基本演算,データ型の基礎,ベクトルの操作方法
 
Mat fin
Mat finMat fin
Mat fin
 
إجراءات قائمة الشحن
إجراءات قائمة الشحنإجراءات قائمة الشحن
إجراءات قائمة الشحن
 
音響信号に対する異常音検知技術と応用
音響信号に対する異常音検知技術と応用音響信号に対する異常音検知技術と応用
音響信号に対する異常音検知技術と応用
 
الإستاتيكا
الإستاتيكاالإستاتيكا
الإستاتيكا
 
機械学習と自動微分
機械学習と自動微分機械学習と自動微分
機械学習と自動微分
 

More from MasanoriSuganuma

0から理解するニューラルネットアーキテクチャサーチ(NAS)
0から理解するニューラルネットアーキテクチャサーチ(NAS)0から理解するニューラルネットアーキテクチャサーチ(NAS)
0から理解するニューラルネットアーキテクチャサーチ(NAS)
MasanoriSuganuma
 
RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
RAFT: Recurrent All-Pairs Field Transforms for Optical FlowRAFT: Recurrent All-Pairs Field Transforms for Optical Flow
RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
MasanoriSuganuma
 
高1から始める人工知能(AI)
高1から始める人工知能(AI)高1から始める人工知能(AI)
高1から始める人工知能(AI)
MasanoriSuganuma
 
When NAS Meets Robustness: In Search of Robust Architectures against Adversar...
When NAS Meets Robustness:In Search of Robust Architectures againstAdversar...When NAS Meets Robustness:In Search of Robust Architectures againstAdversar...
When NAS Meets Robustness: In Search of Robust Architectures against Adversar...
MasanoriSuganuma
 
When NAS Meets Robustness: In Search of Robust Architectures against Adversar...
When NAS Meets Robustness:In Search of Robust Architectures againstAdversar...When NAS Meets Robustness:In Search of Robust Architectures againstAdversar...
When NAS Meets Robustness: In Search of Robust Architectures against Adversar...
MasanoriSuganuma
 
CNNの構造最適化手法について
CNNの構造最適化手法についてCNNの構造最適化手法について
CNNの構造最適化手法について
MasanoriSuganuma
 
CNNの構造最適化手法(第3回3D勉強会)
CNNの構造最適化手法(第3回3D勉強会)CNNの構造最適化手法(第3回3D勉強会)
CNNの構造最適化手法(第3回3D勉強会)
MasanoriSuganuma
 

More from MasanoriSuganuma (7)

0から理解するニューラルネットアーキテクチャサーチ(NAS)
0から理解するニューラルネットアーキテクチャサーチ(NAS)0から理解するニューラルネットアーキテクチャサーチ(NAS)
0から理解するニューラルネットアーキテクチャサーチ(NAS)
 
RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
RAFT: Recurrent All-Pairs Field Transforms for Optical FlowRAFT: Recurrent All-Pairs Field Transforms for Optical Flow
RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
 
高1から始める人工知能(AI)
高1から始める人工知能(AI)高1から始める人工知能(AI)
高1から始める人工知能(AI)
 
When NAS Meets Robustness: In Search of Robust Architectures against Adversar...
When NAS Meets Robustness:In Search of Robust Architectures againstAdversar...When NAS Meets Robustness:In Search of Robust Architectures againstAdversar...
When NAS Meets Robustness: In Search of Robust Architectures against Adversar...
 
When NAS Meets Robustness: In Search of Robust Architectures against Adversar...
When NAS Meets Robustness:In Search of Robust Architectures againstAdversar...When NAS Meets Robustness:In Search of Robust Architectures againstAdversar...
When NAS Meets Robustness: In Search of Robust Architectures against Adversar...
 
CNNの構造最適化手法について
CNNの構造最適化手法についてCNNの構造最適化手法について
CNNの構造最適化手法について
 
CNNの構造最適化手法(第3回3D勉強会)
CNNの構造最適化手法(第3回3D勉強会)CNNの構造最適化手法(第3回3D勉強会)
CNNの構造最適化手法(第3回3D勉強会)
 

Recently uploaded

Le nuove frontiere dell'AI nell'RPA con UiPath Autopilot™
Le nuove frontiere dell'AI nell'RPA con UiPath Autopilot™Le nuove frontiere dell'AI nell'RPA con UiPath Autopilot™
Le nuove frontiere dell'AI nell'RPA con UiPath Autopilot™
UiPathCommunity
 
Observability Concepts EVERY Developer Should Know -- DeveloperWeek Europe.pdf
Observability Concepts EVERY Developer Should Know -- DeveloperWeek Europe.pdfObservability Concepts EVERY Developer Should Know -- DeveloperWeek Europe.pdf
Observability Concepts EVERY Developer Should Know -- DeveloperWeek Europe.pdf
Paige Cruz
 
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...
Ramesh Iyer
 
Secstrike : Reverse Engineering & Pwnable tools for CTF.pptx
Secstrike : Reverse Engineering & Pwnable tools for CTF.pptxSecstrike : Reverse Engineering & Pwnable tools for CTF.pptx
Secstrike : Reverse Engineering & Pwnable tools for CTF.pptx
nkrafacyberclub
 
Assuring Contact Center Experiences for Your Customers With ThousandEyes
Assuring Contact Center Experiences for Your Customers With ThousandEyesAssuring Contact Center Experiences for Your Customers With ThousandEyes
Assuring Contact Center Experiences for Your Customers With ThousandEyes
ThousandEyes
 
Quantum Computing: Current Landscape and the Future Role of APIs
Quantum Computing: Current Landscape and the Future Role of APIsQuantum Computing: Current Landscape and the Future Role of APIs
Quantum Computing: Current Landscape and the Future Role of APIs
Vlad Stirbu
 
Free Complete Python - A step towards Data Science
Free Complete Python - A step towards Data ScienceFree Complete Python - A step towards Data Science
Free Complete Python - A step towards Data Science
RinaMondal9
 
Welocme to ViralQR, your best QR code generator.
Welocme to ViralQR, your best QR code generator.Welocme to ViralQR, your best QR code generator.
Welocme to ViralQR, your best QR code generator.
ViralQR
 
Key Trends Shaping the Future of Infrastructure.pdf
Key Trends Shaping the Future of Infrastructure.pdfKey Trends Shaping the Future of Infrastructure.pdf
Key Trends Shaping the Future of Infrastructure.pdf
Cheryl Hung
 
Generative AI Deep Dive: Advancing from Proof of Concept to Production
Generative AI Deep Dive: Advancing from Proof of Concept to ProductionGenerative AI Deep Dive: Advancing from Proof of Concept to Production
Generative AI Deep Dive: Advancing from Proof of Concept to Production
Aggregage
 
Bits & Pixels using AI for Good.........
Bits & Pixels using AI for Good.........Bits & Pixels using AI for Good.........
Bits & Pixels using AI for Good.........
Alison B. Lowndes
 
PHP Frameworks: I want to break free (IPC Berlin 2024)
PHP Frameworks: I want to break free (IPC Berlin 2024)PHP Frameworks: I want to break free (IPC Berlin 2024)
PHP Frameworks: I want to break free (IPC Berlin 2024)
Ralf Eggert
 
Elevating Tactical DDD Patterns Through Object Calisthenics
Elevating Tactical DDD Patterns Through Object CalisthenicsElevating Tactical DDD Patterns Through Object Calisthenics
Elevating Tactical DDD Patterns Through Object Calisthenics
Dorra BARTAGUIZ
 
Elizabeth Buie - Older adults: Are we really designing for our future selves?
Elizabeth Buie - Older adults: Are we really designing for our future selves?Elizabeth Buie - Older adults: Are we really designing for our future selves?
Elizabeth Buie - Older adults: Are we really designing for our future selves?
Nexer Digital
 
GraphRAG is All You need? LLM & Knowledge Graph
GraphRAG is All You need? LLM & Knowledge GraphGraphRAG is All You need? LLM & Knowledge Graph
GraphRAG is All You need? LLM & Knowledge Graph
Guy Korland
 
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdfFIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance
 
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
Product School
 
Assure Contact Center Experiences for Your Customers With ThousandEyes
Assure Contact Center Experiences for Your Customers With ThousandEyesAssure Contact Center Experiences for Your Customers With ThousandEyes
Assure Contact Center Experiences for Your Customers With ThousandEyes
ThousandEyes
 
By Design, not by Accident - Agile Venture Bolzano 2024
By Design, not by Accident - Agile Venture Bolzano 2024By Design, not by Accident - Agile Venture Bolzano 2024
By Design, not by Accident - Agile Venture Bolzano 2024
Pierluigi Pugliese
 
Accelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish CachingAccelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish Caching
Thijs Feryn
 

Recently uploaded (20)

Le nuove frontiere dell'AI nell'RPA con UiPath Autopilot™
Le nuove frontiere dell'AI nell'RPA con UiPath Autopilot™Le nuove frontiere dell'AI nell'RPA con UiPath Autopilot™
Le nuove frontiere dell'AI nell'RPA con UiPath Autopilot™
 
Observability Concepts EVERY Developer Should Know -- DeveloperWeek Europe.pdf
Observability Concepts EVERY Developer Should Know -- DeveloperWeek Europe.pdfObservability Concepts EVERY Developer Should Know -- DeveloperWeek Europe.pdf
Observability Concepts EVERY Developer Should Know -- DeveloperWeek Europe.pdf
 
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...
 
Secstrike : Reverse Engineering & Pwnable tools for CTF.pptx
Secstrike : Reverse Engineering & Pwnable tools for CTF.pptxSecstrike : Reverse Engineering & Pwnable tools for CTF.pptx
Secstrike : Reverse Engineering & Pwnable tools for CTF.pptx
 
Assuring Contact Center Experiences for Your Customers With ThousandEyes
Assuring Contact Center Experiences for Your Customers With ThousandEyesAssuring Contact Center Experiences for Your Customers With ThousandEyes
Assuring Contact Center Experiences for Your Customers With ThousandEyes
 
Quantum Computing: Current Landscape and the Future Role of APIs
Quantum Computing: Current Landscape and the Future Role of APIsQuantum Computing: Current Landscape and the Future Role of APIs
Quantum Computing: Current Landscape and the Future Role of APIs
 
Free Complete Python - A step towards Data Science
Free Complete Python - A step towards Data ScienceFree Complete Python - A step towards Data Science
Free Complete Python - A step towards Data Science
 
Welocme to ViralQR, your best QR code generator.
Welocme to ViralQR, your best QR code generator.Welocme to ViralQR, your best QR code generator.
Welocme to ViralQR, your best QR code generator.
 
Key Trends Shaping the Future of Infrastructure.pdf
Key Trends Shaping the Future of Infrastructure.pdfKey Trends Shaping the Future of Infrastructure.pdf
Key Trends Shaping the Future of Infrastructure.pdf
 
Generative AI Deep Dive: Advancing from Proof of Concept to Production
Generative AI Deep Dive: Advancing from Proof of Concept to ProductionGenerative AI Deep Dive: Advancing from Proof of Concept to Production
Generative AI Deep Dive: Advancing from Proof of Concept to Production
 
Bits & Pixels using AI for Good.........
Bits & Pixels using AI for Good.........Bits & Pixels using AI for Good.........
Bits & Pixels using AI for Good.........
 
PHP Frameworks: I want to break free (IPC Berlin 2024)
PHP Frameworks: I want to break free (IPC Berlin 2024)PHP Frameworks: I want to break free (IPC Berlin 2024)
PHP Frameworks: I want to break free (IPC Berlin 2024)
 
Elevating Tactical DDD Patterns Through Object Calisthenics
Elevating Tactical DDD Patterns Through Object CalisthenicsElevating Tactical DDD Patterns Through Object Calisthenics
Elevating Tactical DDD Patterns Through Object Calisthenics
 
Elizabeth Buie - Older adults: Are we really designing for our future selves?
Elizabeth Buie - Older adults: Are we really designing for our future selves?Elizabeth Buie - Older adults: Are we really designing for our future selves?
Elizabeth Buie - Older adults: Are we really designing for our future selves?
 
GraphRAG is All You need? LLM & Knowledge Graph
GraphRAG is All You need? LLM & Knowledge GraphGraphRAG is All You need? LLM & Knowledge Graph
GraphRAG is All You need? LLM & Knowledge Graph
 
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdfFIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
 
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
 
Assure Contact Center Experiences for Your Customers With ThousandEyes
Assure Contact Center Experiences for Your Customers With ThousandEyesAssure Contact Center Experiences for Your Customers With ThousandEyes
Assure Contact Center Experiences for Your Customers With ThousandEyes
 
By Design, not by Accident - Agile Venture Bolzano 2024
By Design, not by Accident - Agile Venture Bolzano 2024By Design, not by Accident - Agile Venture Bolzano 2024
By Design, not by Accident - Agile Venture Bolzano 2024
 
Accelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish CachingAccelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish Caching
 

Attention-Based Adaptive Selection of Operations for Image Restoration in the Presence of Unknown Combined Distortions [CVPR 2019 読み会]

  • 1. CC6 C 2B6 2 C D6 6 6 C 6A2C B A 2 6 .6BC A2C C96 -A6B6 6 0 3 6 BC AC B I 1-.
  • 2. • • ) [ V • A O LIR +1 ]LIDB E LC • ( A O B E LC • = + . G M 8 7 )9 P . G M 81 2 (9 P . 80 , 9 @
  • 3. - - - - - - - - - - - - - • , 3 3 • A Input Ours Ground truth Mix Raindrop Blur Noise JPEG SE R T UA PXC I MKLNE O
  • 6. ,,K QHN • . 6 • VRP ILM 2 0 2 0 + C 2 C 0 , 219 8
  • 7. C C NO P , ILE A K D I EI DI E I ED E C IE I ED • D DE 0 D C D 0 D 1 A M A A I D • 1E EA 2D K IM , • 9 E A 7 • 5
  • 8. 2 8 810 1 , 1 2 8 810 881 82 1 , 1
  • 10. D A F • 3 03 • 3 • 0 3 1 3 1 -
  • 11. • 2 ( ) 6 • 2 ( ) 12 3×3
  • 12. • 3 - )2 3×3 • ( 1 ) (
  • 14. AC + 4 18C 7E 8 8 F 8 CC 01 . ,E 2AE88H8 7 8 F C 01
  • 15. ! ∈ ℝ$ • 5%&'( ∈ ℝ)×+×$ - ! ∈ ℝ$ • 1
  • 16. !" = [%&" ' , … , %&" * ] • ,- ∈ ℝ0×2, ,3 ∈ ℝ|*|×05" 6" = ,78(,'6") %&" ; = exp(5"(6")) ∑ exp(5"(6")) = exp(&" ; ) ∑;@' |*| exp(&" ; ) → B C 6 1
  • 17. ! "# • $%&'(')*% +,-. 1×1 • $%&'(')*% +,-. 3×3 • $%&'(')*% +,-. 5×5 • $%&'(')*% +,-. 7×7 • $%&'(')*% 45*'6%4 +,-. 3×3 (( = 2) • $%&'(')*% 45*'6%4 +,-. 5×5 (( = 2) • $%&'(')*% 45*'6%4 +,-. 7×7 (( = 2) • '.%(';% &,,*5-; 3×3
  • 19. • - = 1 • 4 9 = • = Feature maps !×#×$ %&'( )& *+,-( Feature maps !×#×$ %& Feature maps !×#×$ %&-,'( … OperationLayerAttentionLayer … Concat 1×1conv. /& OperationLayer Concat 1×1conv. *+,-,
  • 20. • - -- - 2 0 - Feature maps !×#×$ %&'( )& *+,-( Feature maps !×#×$ %& Feature maps !×#×$ %&-,'( … OperationLayerAttentionLayer … Concat 1×1conv. /& OperationLayer Concat 1×1conv. *+,-,
  • 21. • 2 • 1 3 1 23 1 3 , • 2 + • 1 3 1 23 1 3 , • 2 • 1 1 1 3 1 23 1 3 , 3 312
  • 22. , • D : 8 : : • 63×63 C I: PYV R[V K V ] • : + C 8 : : : + C 8 : • 2 b • a 4 5 9 ] , , : - :9 4 5 3 , 8 : : 9::C : 8: : : . ,201
  • 23. 02.1 22 D IK 16 8 76 3 , 54 4 6 EP PE 6I D 6MDE E :ESE E 6E IC 8:79 ::36 8:79 ::36 8:79 ::36 1L077 ) ) ) ) ) )- ( 95 9EP M E , ( () )) ) )- ) 4 L E ETMLD RPPI L DELMIPE . 9EPIDR E LIL M DEEN CLL M IK E DELMIPIL 3222 LP C IMLP ML 3K E 8 MCEPPIL . ( )) 4 R E 0 IL MM C IL M IK E EP M IML BT DEEN EIL M CEKEL E LIL 3L 0 89 , • v tnr 077edf • v uo 8 E ILED 077cighm]aYZp VZs bl [ u DIP M IML l
  • 24. 1 4 - 4 2- 2 42 4
  • 25. PS IG 5 CD 5D R G PSV JL O • 1(3)( 0) 5 • 33 A D 5 • ,5ECC 5 C ,5ECC 5 E 1 , 7 A CC 5 D 57DC ( CD D C 2 CD 0E D 7D
  • 26. ] V [cedf kgn 73 7 LMIK ob a • h mi P I K LMIK M I Yl aYR 182785 6 1 76 .41 01 1 376 /6 KI K I M IMM L K (. ) ) ) , . ( ( ( . )- ,) . ,- % % , . % ) ), - % (, (. - ( , ( . , ) ( ) ,( ) ,( - ) ) , , . ( % % % M C K IP M I CIKL KLI M LC LIA MK MO (( , . ) ( )( . ) )) - ( () ) - - ( ) , ,- - - ( % . , , - % - , % ) - % ) - ( (. ( ( ) , )) , () , )( ) ( (( - ). - . - ) - - - . . , , ( , . ( ) ( ( , . % % % 850 , 89 73 7 LMIK P I K LMIK M I 89 73 7 LMIK P I K LMIK M I 4 MCI 2 9 M 0K AM MII C AIK K LMIK M I K AIK M K 1 0867 -
  • 28. I JE A G • .2 2 2A5 0 1 IRQ • 128×128 5 I • 2 A5 2 385A 5A A5 2 385A • ( , ) I P ) 0 1 . 2 5 2 5 C5 5 5 2 C5 2 C5 A2 2 5 D : 2 5 C2 2 A 5 2 5 ,. 52 A 5
  • 29. • S - . 2 / • - . 2 >9 - - 9 D • - - - - • a - G A
  • 30.
  • 31. • RE N g Ga e i J • 1 h RE N • 33 3 33 3 P d J 1 E B P G
  • 32. : 9 58 2 199A () (- . -. 4429 TA) () . .( 3 103A ( - .(+ 9 ( 4 351 04 89 48 : 9 58 2 199A ( ) -- , 8 N9 TA+ ( - -,- 9 . 4. 84 1 9168 .5 .9 42. 9168 .5 SK IM FKSTPRTKP h] bifkald [c eg A) 6 C I 4 F T P CRF C CST C F M KDM SPM TKP PR E DCS F KNCI F PKSK I 5333 RC SCETKP S P 5NCI :RPE SSK I , . )+ -Z)+ - A 8 IC NC 3 MPKTK I T PT TKCM P STC FCRF EP VPM TKP CM C TP EPF RS PR KNCI R STPRCTKP DY VPM TKP CRY S CRE 5 5187 - A+ CK 8 N T 0 RSKST T N NPRY T PRL PR KNCI R STPRCTKP 5 511 ,
  • 33. • 2CC C a • t gec gecVy ] R • C V s • kbfapKy 0 1 Y I- V r R • u 2CC C a R - 2 3 Y]K ldm n hio c P [ VN L[ 0 1 C 2 2 C 8 2 C 2 28 C 2C 3 C 2 8 .,-
  • 34. • • - - • e r p • D V I • - 43 l I n • - d 4 - ? e o m • • b g h d 4 a - - i • - -