SlideShare a Scribd company logo

Project_Automatic Photo Classification Web Service

•
•
Hyo jeong Lee
Hyo jeong Lee

The document describes a web application called Girlclash that provides automatic photo classification. It uses a convolutional neural network model trained on a dataset of over 40,000 images across 5 categories. The system is built with Django and TensorFlow, and uses tools like CNNs, max pooling and ReLU activation to achieve over 85% accuracy. Areas for further improvement include reducing misclassifications, adding download and member features, and expanding to apps and social media recommendations.

1 of 43
Download to read offline
17-03-31 SookmyungGo Girlclash 1 Girlclash SookmyungGo Girlclash Automatic Photo Classification Web Service 숙명고팀 | 방누리 신아영 이효정 지도교수 | 최영우 교수님 G
17-03-31 SookmyungGo Girlclash 2 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
17-03-31 SookmyungGo Girlclash 3 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
17-03-31 SookmyungGo Girlclash 4 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
17-03-31 SookmyungGo Girlclash 5 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
17-03-31 SookmyungGo Girlclash 6 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
17-03-31 SookmyungGo Girlclash 7 What is Girlclash?
17-03-31 SookmyungGo Girlclash 8 What is Girlclash? Deep Learning SMGO
17-03-31 SookmyungGo Girlclash 9 What is Girlclash?
17-03-31 SookmyungGo Girlclash 10 What is Girlclash? Automatic Photo Classification Web Service
17-03-31 SookmyungGo Girlclash 11 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
17-03-31 SookmyungGo Girlclash 12 How does it work? – Tools DJANGO TENSORFLOW CNN MODEL Python Web Google’s Machine Learning Library Convolutional Neural Network for Image Process
17-03-31 SookmyungGo Girlclash 13 How does it work? – Tools DJANGO Python Web OBJECT RELATIONAL MAPPING •다양한 DB 지원 •유연성 •다양한 내장 함수 BACKEND CONSOLE •콘솔 상 DB 관리 가능 •NoSQL MODULE •프로젝트-모듈 분리 •전체 개발시간 단축
17-03-31 SookmyungGo Girlclash 14 How does it work? – Tools TENSORFLOW Google’s Machine Learning Library STRENGTH •Tensorboard로 구조도 제공 •사용자 정의 신경망 가능 •복잡한 연산 자동 수행 DEFINITION •Tensor란? 동적 다차원 데이터 배열 •노드 = 연산 엣지 = 데이터 전달 데이터 = Tensor GRAPH OPERATION
17-03-31 SookmyungGo Girlclash 15 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process LENET 1998 ALEXNET 2012 •Handwritten digits (Grayscale) •Average pooling Layer •Sigmoid function •RGB images •Max pooling layer •ReLU function
17-03-31 SookmyungGo Girlclash 16 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process LENET 1998 ALEXNET 2012 •Handwritten digits (Grayscale) •Average pooling Layer •Sigmoid function •RGB images •Max pooling layer •ReLU function
17-03-31 SookmyungGo Girlclash 17 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •Dropout A B C
17-03-31 SookmyungGo Girlclash 18 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •Dropout
17-03-31 SookmyungGo Girlclash 19 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012
17-03-31 SookmyungGo Girlclash 20 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •Dropout
17-03-31 SookmyungGo Girlclash 21 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •Dropout 50 100 Sigmoid ReLU 정확도(%) 활성화함수 활성화 함수에 따른 정확도(%)
17-03-31 SookmyungGo Girlclash 22 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •Dropout
17-03-31 SookmyungGo Girlclash 23 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •LRN & Dropout
17-03-31 SookmyungGo Girlclash 24 참고. CNN 학습 모델의 형태 (tensorboard)
17-03-31 SookmyungGo Girlclash 25 참고. CNN 학습 모델의 형태 (diagram)
17-03-31 SookmyungGo Girlclash 26 How does it work? – Machine Learning Environment Learning Conditions CNNs Misclassification CPU GPU LeNet AlexNet Time Rate Dataset Solution
17-03-31 SookmyungGo Girlclash 27 How does it work? – Machine Learning ENVIRONMENT 60 3 CPU GPU 시간(분) 환경 학습 환경에 따른 소요 시간(분) CPU GPU • 데이터셋 : 약 40,000장 • 신경망 : LeNet • 학습 횟수 : 8,500회 GTX1070 with
17-03-31 SookmyungGo Girlclash 28 How does it work? – Machine Learning • 5-class Dataset Faces 9,958 Pets 12,474 Food 9,866 Nature 8,703 Fashion 11,786 LEARNING CONDITION - DATASET Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Dataset Rate Time
17-03-31 SookmyungGo Girlclash 29 How does it work? – Machine Learning • Learning Rate LEARNING CONDITION – LEARNING RATE Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Dataset Rate Time45.5 41.5 52.8 45.8 46.8 0.001 0.003 0.005 0.007 0.01 정확도(%) 학습율 학습율에 따른 정확도(%) • 데이터셋 : 약 40,000장 • 신경망 : AlexNet • 학습 횟수 : 5,000회 ~
17-03-31 SookmyungGo Girlclash 30 How does it work? – Machine Learning • Learning Time LEARNING CONDITION – LEARNING TIME Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Dataset Rate Time71.8 93.7 93.7 85.9 96.3 97.8 8,500회 100,000회 500,000회 정확도(%) 학습횟수 학습 횟수에 따른 정확도(%) train test • 데이터셋 : 약 40,000장 • 신경망 : Alexnet
17-03-31 SookmyungGo Girlclash 31 How does it work? – Machine Learning • LeNet • AlexNet CNNS Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum 60 3 CPU GPU 정확도(%) 신경망 신경망에 따른 정확도(%) 열1 Series 1 • 데이터셋 : 약 40,000장 • 학습 횟수 : 8,500회 66.8 71.8 72 85.9 LeNet AlexNet 정확도(%) 신경망 신경망에 따른 정확도(%) train test
17-03-31 SookmyungGo Girlclash 32 How does it work? – Machine Learning 1. 오분류 해결 방법 A. Prediction : 예측치 < 0.5인 이미지를 Etc로 분류 B. Garbage : Garbage Class를 포함하여 학습 MISCLASSIFICATION Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Solution Test case Soccer ball Umbrella Airplane Car 개수 63 75 800 8144 Garbage class 내 포함 여부 미포함 포함 미포함이지만 유사함
17-03-31 SookmyungGo Girlclash 33 How does it work? – Machine Learning 1. 오분류 해결 방법 A. Prediction : 예측치 < 0.5인 이미지를 Etc로 분류 B. Garbage : Garbage Class를 포함하여 학습 MISCLASSIFICATION Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Solution 1 2 3 4 5 6 (garbage)
17-03-31 SookmyungGo Girlclash 34 How does it work? – Machine Learning 1. 오분류 해결 방법 A. Prediction : 예측치 < 0.5인 이미지를 Etc로 분류 B. Garbage : Garbage Class를 포함하여 학습 2. Test case MISCLASSIFICATION Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Solution Test case Soccer ball Umbrella Airplane Car 개수 63 75 800 8144 Garbage class 내 포함 여부 미포함 포함 미포함이지만 유사함
17-03-31 SookmyungGo Girlclash 35 How does it work? – Machine Learning MISCLASSIFICATION Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Solution 31.3 35.9 44 42.7 94.9 1.5 36.3 35.5 Garbage Prediction 성공률(%) 오분류해결방법 해결 방법에 따른 분류 성공률(%) Car Airplane Umbrella Soccer ball • 학습 횟수 : 100,000회 • 신경망 : AlexnetA B
17-03-31 SookmyungGo Girlclash 36 How does it work? – Machine Learning MISCLASSIFICATION Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Solution 1 2 3 4 5 6 (garbage) B
17-03-31 SookmyungGo Girlclash 37 How does it work? – Web Application 시연 영상 및 설명
17-03-31 SookmyungGo Girlclash 38 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
17-03-31 SookmyungGo Girlclash 39 Things to improve 압축 파일 다운로드 기능 추가 오분류 개선 회원 관리 기능 앱 개발
17-03-31 SookmyungGo Girlclash 40 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
17-03-31 SookmyungGo Girlclash 41 What we expect SNS로 확장 사진 추천 기능 추가
17-03-31 SookmyungGo Girlclash 42 Thank you for your attention! Do you have any questions? 방누리 신아영 이효정 웹 프론트엔드 /백엔드 데이터셋 수집 /웹 프론트엔드 딥러닝 /웹과 분류코드 연결
17-03-31 SookmyungGo Girlclash 43 • CNN 학습 모델 - http://dsmoon.tistory.com/entry/TensorFlow-tutorial-Convolutional-Neural-Networks • LeNet과 AlexNet - 밑바닥부터 시작하는 딥러닝 (한빛미디어, 사이토 고키, 2017) • CNN - http://hamait.tistory.com/535 • 활성화 함수 - 모두의 딥러닝 (김성훈) • 드롭아웃 - http://bcho.tistory.com/tag/alexnet • 풀링 - https://tensorflow.blog/%ED%95%B4%EC%BB%A4%EC%97%90%EA%B2%8C- %EC%A0%84%ED%95%B4%EB%93%A4%EC%9D%80- %EB%A8%B8%EC%8B%A0%EB%9F%AC%EB%8B%9D-4/ - https://www.embedded-vision.com/platinum-members/cadence/embedded-vision- training/documents/pages/neuralnetworksimagerecognition - http://sanghyukchun.github.io/75/ 출처

Recommended

Semi-Supervised Insight Generation from Petabyte Scale Text Data
Semi-Supervised Insight Generation from Petabyte Scale Text DataSemi-Supervised Insight Generation from Petabyte Scale Text Data
Semi-Supervised Insight Generation from Petabyte Scale Text Data
Tech Triveni
 
Using Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning ModelsUsing Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning Models
Scott Clark
 
Using Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning ModelsUsing Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning Models
SigOpt
 
Using SigOpt to Tune Deep Learning Models with Nervana Cloud
Using SigOpt to Tune Deep Learning Models with Nervana CloudUsing SigOpt to Tune Deep Learning Models with Nervana Cloud
Using SigOpt to Tune Deep Learning Models with Nervana Cloud
SigOpt
 
DiscoGAN - Learning to Discover Cross-Domain Relations with Generative Advers...
DiscoGAN - Learning to Discover Cross-Domain Relations with Generative Advers...DiscoGAN - Learning to Discover Cross-Domain Relations with Generative Advers...
DiscoGAN - Learning to Discover Cross-Domain Relations with Generative Advers...
Taeksoo Kim
 
Haystack 2019 - Search Logs + Machine Learning = Auto-Tagging Inventory - Joh...
Haystack 2019 - Search Logs + Machine Learning = Auto-Tagging Inventory - Joh...Haystack 2019 - Search Logs + Machine Learning = Auto-Tagging Inventory - Joh...
Haystack 2019 - Search Logs + Machine Learning = Auto-Tagging Inventory - Joh...
OpenSource Connections
 
Role of Data Science in eCommerce
Role of Data Science in eCommerceRole of Data Science in eCommerce
Role of Data Science in eCommerce
ManojKumar Rangasamy Kannadasan
 
Extending the breadth and depth of interaction using gamification
Extending the breadth and depth of interaction using gamificationExtending the breadth and depth of interaction using gamification
Extending the breadth and depth of interaction using gamification
Malcolm Murray
 

Recommended

Semi-Supervised Insight Generation from Petabyte Scale Text Data
Semi-Supervised Insight Generation from Petabyte Scale Text DataSemi-Supervised Insight Generation from Petabyte Scale Text Data
Semi-Supervised Insight Generation from Petabyte Scale Text Data
Tech Triveni
 
Using Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning ModelsUsing Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning Models
Scott Clark
 
Using Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning ModelsUsing Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning Models
SigOpt
 
Using SigOpt to Tune Deep Learning Models with Nervana Cloud
Using SigOpt to Tune Deep Learning Models with Nervana CloudUsing SigOpt to Tune Deep Learning Models with Nervana Cloud
Using SigOpt to Tune Deep Learning Models with Nervana Cloud
SigOpt
 
DiscoGAN - Learning to Discover Cross-Domain Relations with Generative Advers...
DiscoGAN - Learning to Discover Cross-Domain Relations with Generative Advers...DiscoGAN - Learning to Discover Cross-Domain Relations with Generative Advers...
DiscoGAN - Learning to Discover Cross-Domain Relations with Generative Advers...
Taeksoo Kim
 
Haystack 2019 - Search Logs + Machine Learning = Auto-Tagging Inventory - Joh...
Haystack 2019 - Search Logs + Machine Learning = Auto-Tagging Inventory - Joh...Haystack 2019 - Search Logs + Machine Learning = Auto-Tagging Inventory - Joh...
Haystack 2019 - Search Logs + Machine Learning = Auto-Tagging Inventory - Joh...
OpenSource Connections
 
Role of Data Science in eCommerce
Role of Data Science in eCommerceRole of Data Science in eCommerce
Role of Data Science in eCommerce
ManojKumar Rangasamy Kannadasan
 
Extending the breadth and depth of interaction using gamification
Extending the breadth and depth of interaction using gamificationExtending the breadth and depth of interaction using gamification
Extending the breadth and depth of interaction using gamification
Malcolm Murray
 
Designing for garbage collection
Designing for garbage collectionDesigning for garbage collection
Designing for garbage collectionGregg Donovan
 
Dynamic Search and Beyond
Dynamic Search and BeyondDynamic Search and Beyond
Dynamic Search and Beyond
Grace Hui Yang
 
The Frontier of Deep Learning in 2020 and Beyond
The Frontier of Deep Learning in 2020 and BeyondThe Frontier of Deep Learning in 2020 and Beyond
The Frontier of Deep Learning in 2020 and Beyond
NUS-ISS
 
Supervised learning
Supervised learningSupervised learning
Supervised learning
O. R. Kumaran
 
How DeepMind Mastered The Game Of Go
How DeepMind Mastered The Game Of GoHow DeepMind Mastered The Game Of Go
How DeepMind Mastered The Game Of Go
Tim Riser
 
Deep Convolutional GANs - meaning of latent space
Deep Convolutional GANs - meaning of latent spaceDeep Convolutional GANs - meaning of latent space
Deep Convolutional GANs - meaning of latent space
Hansol Kang
 
Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...
Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...
Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...
MLconf
 
MLSD18. Ensembles, Logistic Regression, Deepnets
MLSD18. Ensembles, Logistic Regression, DeepnetsMLSD18. Ensembles, Logistic Regression, Deepnets
MLSD18. Ensembles, Logistic Regression, Deepnets
BigML, Inc
 
Db migrations equal pain
Db migrations equal painDb migrations equal pain
Db migrations equal pain
Eugen Oskin
 
ML Study Jams - Session 3.pptx
ML Study Jams - Session 3.pptxML Study Jams - Session 3.pptx
ML Study Jams - Session 3.pptx
MayankChadha14
 
Approximate "Now" is Better Than Accurate "Later"
Approximate "Now" is Better Than Accurate "Later"Approximate "Now" is Better Than Accurate "Later"
Approximate "Now" is Better Than Accurate "Later"
NUS-ISS
 
Automatic Image Cropping - A journey from a Master Thesis to Production
Automatic Image Cropping - A journey from a Master Thesis to ProductionAutomatic Image Cropping - A journey from a Master Thesis to Production
Automatic Image Cropping - A journey from a Master Thesis to Production
Alexey Grigorev
 
Teaching Your Computer To Play Video Games
Teaching Your Computer To Play Video GamesTeaching Your Computer To Play Video Games
Teaching Your Computer To Play Video Games
ehrenbrav
 
Leveraging AI & ML to Automoate Repetitive Tasks
Leveraging AI & ML to Automoate Repetitive TasksLeveraging AI & ML to Automoate Repetitive Tasks
Leveraging AI & ML to Automoate Repetitive Tasks
SabrinaBandel1
 
Model based rl
Model based rlModel based rl
Model based rl
Seolhokim
 
Understanding computer vision with Deep Learning
Understanding computer vision with Deep LearningUnderstanding computer vision with Deep Learning
Understanding computer vision with Deep Learning
CloudxLab
 
Understanding computer vision with Deep Learning
Understanding computer vision with Deep LearningUnderstanding computer vision with Deep Learning
Understanding computer vision with Deep Learning
knowbigdata
 
Understanding computer vision with Deep Learning
Understanding computer vision with Deep LearningUnderstanding computer vision with Deep Learning
Understanding computer vision with Deep Learning
ShubhWadekar
 
Ropossum: A Game That Generates Itself
Ropossum: A Game That Generates ItselfRopossum: A Game That Generates Itself
Ropossum: A Game That Generates Itself
Mohammad Shaker
 
Model selection and tuning at scale
Model selection and tuning at scaleModel selection and tuning at scale
Model selection and tuning at scale
Owen Zhang
 
Progress_190118
Progress_190118Progress_190118
Progress_190118
Hyo jeong Lee
 
Progress_190130
Progress_190130Progress_190130
Progress_190130
Hyo jeong Lee
 

More Related Content

Similar to Project_Automatic Photo Classification Web Service

Designing for garbage collection
Designing for garbage collectionDesigning for garbage collection
Designing for garbage collectionGregg Donovan
 
Dynamic Search and Beyond
Dynamic Search and BeyondDynamic Search and Beyond
Dynamic Search and Beyond
Grace Hui Yang
 
The Frontier of Deep Learning in 2020 and Beyond
The Frontier of Deep Learning in 2020 and BeyondThe Frontier of Deep Learning in 2020 and Beyond
The Frontier of Deep Learning in 2020 and Beyond
NUS-ISS
 
Supervised learning
Supervised learningSupervised learning
Supervised learning
O. R. Kumaran
 
How DeepMind Mastered The Game Of Go
How DeepMind Mastered The Game Of GoHow DeepMind Mastered The Game Of Go
How DeepMind Mastered The Game Of Go
Tim Riser
 
Deep Convolutional GANs - meaning of latent space
Deep Convolutional GANs - meaning of latent spaceDeep Convolutional GANs - meaning of latent space
Deep Convolutional GANs - meaning of latent space
Hansol Kang
 
Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...
Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...
Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...
MLconf
 
MLSD18. Ensembles, Logistic Regression, Deepnets
MLSD18. Ensembles, Logistic Regression, DeepnetsMLSD18. Ensembles, Logistic Regression, Deepnets
MLSD18. Ensembles, Logistic Regression, Deepnets
BigML, Inc
 
Db migrations equal pain
Db migrations equal painDb migrations equal pain
Db migrations equal pain
Eugen Oskin
 
ML Study Jams - Session 3.pptx
ML Study Jams - Session 3.pptxML Study Jams - Session 3.pptx
ML Study Jams - Session 3.pptx
MayankChadha14
 
Approximate "Now" is Better Than Accurate "Later"
Approximate "Now" is Better Than Accurate "Later"Approximate "Now" is Better Than Accurate "Later"
Approximate "Now" is Better Than Accurate "Later"
NUS-ISS
 
Automatic Image Cropping - A journey from a Master Thesis to Production
Automatic Image Cropping - A journey from a Master Thesis to ProductionAutomatic Image Cropping - A journey from a Master Thesis to Production
Automatic Image Cropping - A journey from a Master Thesis to Production
Alexey Grigorev
 
Teaching Your Computer To Play Video Games
Teaching Your Computer To Play Video GamesTeaching Your Computer To Play Video Games
Teaching Your Computer To Play Video Games
ehrenbrav
 
Leveraging AI & ML to Automoate Repetitive Tasks
Leveraging AI & ML to Automoate Repetitive TasksLeveraging AI & ML to Automoate Repetitive Tasks
Leveraging AI & ML to Automoate Repetitive Tasks
SabrinaBandel1
 
Model based rl
Model based rlModel based rl
Model based rl
Seolhokim
 
Understanding computer vision with Deep Learning
Understanding computer vision with Deep LearningUnderstanding computer vision with Deep Learning
Understanding computer vision with Deep Learning
CloudxLab
 
Understanding computer vision with Deep Learning
Understanding computer vision with Deep LearningUnderstanding computer vision with Deep Learning
Understanding computer vision with Deep Learning
knowbigdata
 
Understanding computer vision with Deep Learning
Understanding computer vision with Deep LearningUnderstanding computer vision with Deep Learning
Understanding computer vision with Deep Learning
ShubhWadekar
 
Ropossum: A Game That Generates Itself
Ropossum: A Game That Generates ItselfRopossum: A Game That Generates Itself
Ropossum: A Game That Generates Itself
Mohammad Shaker
 
Model selection and tuning at scale
Model selection and tuning at scaleModel selection and tuning at scale
Model selection and tuning at scale
Owen Zhang
 

Similar to Project_Automatic Photo Classification Web Service (20)

Designing for garbage collection
Designing for garbage collectionDesigning for garbage collection
Designing for garbage collection
 
Dynamic Search and Beyond
Dynamic Search and BeyondDynamic Search and Beyond
Dynamic Search and Beyond
 
The Frontier of Deep Learning in 2020 and Beyond
The Frontier of Deep Learning in 2020 and BeyondThe Frontier of Deep Learning in 2020 and Beyond
The Frontier of Deep Learning in 2020 and Beyond
 
Supervised learning
Supervised learningSupervised learning
Supervised learning
 
How DeepMind Mastered The Game Of Go
How DeepMind Mastered The Game Of GoHow DeepMind Mastered The Game Of Go
How DeepMind Mastered The Game Of Go
 
Deep Convolutional GANs - meaning of latent space
Deep Convolutional GANs - meaning of latent spaceDeep Convolutional GANs - meaning of latent space
Deep Convolutional GANs - meaning of latent space
 
Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...
Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...
Le Song, Assistant Professor, College of Computing, Georgia Institute of Tech...
 
MLSD18. Ensembles, Logistic Regression, Deepnets
MLSD18. Ensembles, Logistic Regression, DeepnetsMLSD18. Ensembles, Logistic Regression, Deepnets
MLSD18. Ensembles, Logistic Regression, Deepnets
 
Db migrations equal pain
Db migrations equal painDb migrations equal pain
Db migrations equal pain
 
ML Study Jams - Session 3.pptx
ML Study Jams - Session 3.pptxML Study Jams - Session 3.pptx
ML Study Jams - Session 3.pptx
 
Approximate "Now" is Better Than Accurate "Later"
Approximate "Now" is Better Than Accurate "Later"Approximate "Now" is Better Than Accurate "Later"
Approximate "Now" is Better Than Accurate "Later"
 
Automatic Image Cropping - A journey from a Master Thesis to Production
Automatic Image Cropping - A journey from a Master Thesis to ProductionAutomatic Image Cropping - A journey from a Master Thesis to Production
Automatic Image Cropping - A journey from a Master Thesis to Production
 
Teaching Your Computer To Play Video Games
Teaching Your Computer To Play Video GamesTeaching Your Computer To Play Video Games
Teaching Your Computer To Play Video Games
 
Leveraging AI & ML to Automoate Repetitive Tasks
Leveraging AI & ML to Automoate Repetitive TasksLeveraging AI & ML to Automoate Repetitive Tasks
Leveraging AI & ML to Automoate Repetitive Tasks
 
Model based rl
Model based rlModel based rl
Model based rl
 
Understanding computer vision with Deep Learning
Understanding computer vision with Deep LearningUnderstanding computer vision with Deep Learning
Understanding computer vision with Deep Learning
 
Understanding computer vision with Deep Learning
Understanding computer vision with Deep LearningUnderstanding computer vision with Deep Learning
Understanding computer vision with Deep Learning
 
Understanding computer vision with Deep Learning
Understanding computer vision with Deep LearningUnderstanding computer vision with Deep Learning
Understanding computer vision with Deep Learning
 
Ropossum: A Game That Generates Itself
Ropossum: A Game That Generates ItselfRopossum: A Game That Generates Itself
Ropossum: A Game That Generates Itself
 
Model selection and tuning at scale
Model selection and tuning at scaleModel selection and tuning at scale
Model selection and tuning at scale
 

More from Hyo jeong Lee

Progress_190118
Progress_190118Progress_190118
Progress_190118
Hyo jeong Lee
 
Progress_190130
Progress_190130Progress_190130
Progress_190130
Hyo jeong Lee
 
Progress_190213
Progress_190213Progress_190213
Progress_190213
Hyo jeong Lee
 
Progress_190412
Progress_190412Progress_190412
Progress_190412
Hyo jeong Lee
 
Progress_190315
Progress_190315Progress_190315
Progress_190315
Hyo jeong Lee
 
Paper_Scalable database logging for multicores
Paper_Scalable database logging for multicoresPaper_Scalable database logging for multicores
Paper_Scalable database logging for multicores
Hyo jeong Lee
 
Paper_An Efficient Garbage Collection in Java Virtual Machine via Swap I/O O...
Paper_An Efficient Garbage Collection in Java Virtual Machine via Swap I/O O...Paper_An Efficient Garbage Collection in Java Virtual Machine via Swap I/O O...
Paper_An Efficient Garbage Collection in Java Virtual Machine via Swap I/O O...
Hyo jeong Lee
 
Paper_Design of Swap-aware Java Virtual Machine Garbage Collector Policy
Paper_Design of Swap-aware Java Virtual Machine Garbage Collector PolicyPaper_Design of Swap-aware Java Virtual Machine Garbage Collector Policy
Paper_Design of Swap-aware Java Virtual Machine Garbage Collector Policy
Hyo jeong Lee
 
Howto_Tensorflow-slim
Howto_Tensorflow-slimHowto_Tensorflow-slim
Howto_Tensorflow-slim
Hyo jeong Lee
 
Howto_Tensorflow+Linear Regression
Howto_Tensorflow+Linear RegressionHowto_Tensorflow+Linear Regression
Howto_Tensorflow+Linear Regression
Hyo jeong Lee
 

More from Hyo jeong Lee (10)

Progress_190118
Progress_190118Progress_190118
Progress_190118
 
Progress_190130
Progress_190130Progress_190130
Progress_190130
 
Progress_190213
Progress_190213Progress_190213
Progress_190213
 
Progress_190412
Progress_190412Progress_190412
Progress_190412
 
Progress_190315
Progress_190315Progress_190315
Progress_190315
 
Paper_Scalable database logging for multicores
Paper_Scalable database logging for multicoresPaper_Scalable database logging for multicores
Paper_Scalable database logging for multicores
 
Paper_An Efficient Garbage Collection in Java Virtual Machine via Swap I/O O...
Paper_An Efficient Garbage Collection in Java Virtual Machine via Swap I/O O...Paper_An Efficient Garbage Collection in Java Virtual Machine via Swap I/O O...
Paper_An Efficient Garbage Collection in Java Virtual Machine via Swap I/O O...
 
Paper_Design of Swap-aware Java Virtual Machine Garbage Collector Policy
Paper_Design of Swap-aware Java Virtual Machine Garbage Collector PolicyPaper_Design of Swap-aware Java Virtual Machine Garbage Collector Policy
Paper_Design of Swap-aware Java Virtual Machine Garbage Collector Policy
 
Howto_Tensorflow-slim
Howto_Tensorflow-slimHowto_Tensorflow-slim
Howto_Tensorflow-slim
 
Howto_Tensorflow+Linear Regression
Howto_Tensorflow+Linear RegressionHowto_Tensorflow+Linear Regression
Howto_Tensorflow+Linear Regression
 

Recently uploaded

Innovating Inference - Remote Triggering of Large Language Models on HPC Clus...
Innovating Inference - Remote Triggering of Large Language Models on HPC Clus...Innovating Inference - Remote Triggering of Large Language Models on HPC Clus...
Innovating Inference - Remote Triggering of Large Language Models on HPC Clus...
Globus
 
Enterprise Resource Planning System in Telangana
Enterprise Resource Planning System in TelanganaEnterprise Resource Planning System in Telangana
Enterprise Resource Planning System in Telangana
NYGGS Automation Suite
 
Beyond Event Sourcing - Embracing CRUD for Wix Platform - Java.IL
Beyond Event Sourcing - Embracing CRUD for Wix Platform - Java.ILBeyond Event Sourcing - Embracing CRUD for Wix Platform - Java.IL
Beyond Event Sourcing - Embracing CRUD for Wix Platform - Java.IL
Natan Silnitsky
 
Understanding Globus Data Transfers with NetSage
Understanding Globus Data Transfers with NetSageUnderstanding Globus Data Transfers with NetSage
Understanding Globus Data Transfers with NetSage
Globus
 
Globus Connect Server Deep Dive - GlobusWorld 2024
Globus Connect Server Deep Dive - GlobusWorld 2024Globus Connect Server Deep Dive - GlobusWorld 2024
Globus Connect Server Deep Dive - GlobusWorld 2024
Globus
 
Webinar: Salesforce Document Management 2.0 - Smarter, Faster, Better
Webinar: Salesforce Document Management 2.0 - Smarter, Faster, BetterWebinar: Salesforce Document Management 2.0 - Smarter, Faster, Better
Webinar: Salesforce Document Management 2.0 - Smarter, Faster, Better
XfilesPro
 
Navigating the Metaverse: A Journey into Virtual Evolution"
Navigating the Metaverse: A Journey into Virtual Evolution"Navigating the Metaverse: A Journey into Virtual Evolution"
Navigating the Metaverse: A Journey into Virtual Evolution"
Donna Lenk
 
Corporate Management | Session 3 of 3 | Tendenci AMS
Corporate Management | Session 3 of 3 | Tendenci AMSCorporate Management | Session 3 of 3 | Tendenci AMS
Corporate Management | Session 3 of 3 | Tendenci AMS
Tendenci - The Open Source AMS (Association Management Software)
 
Accelerate Enterprise Software Engineering with Platformless
Accelerate Enterprise Software Engineering with PlatformlessAccelerate Enterprise Software Engineering with Platformless
Accelerate Enterprise Software Engineering with Platformless
WSO2
 
Using IESVE for Room Loads Analysis - Australia & New Zealand
Using IESVE for Room Loads Analysis - Australia & New ZealandUsing IESVE for Room Loads Analysis - Australia & New Zealand
Using IESVE for Room Loads Analysis - Australia & New Zealand
IES VE
 
Climate Science Flows: Enabling Petabyte-Scale Climate Analysis with the Eart...
Climate Science Flows: Enabling Petabyte-Scale Climate Analysis with the Eart...Climate Science Flows: Enabling Petabyte-Scale Climate Analysis with the Eart...
Climate Science Flows: Enabling Petabyte-Scale Climate Analysis with the Eart...
Globus
 
SOCRadar Research Team: Latest Activities of IntelBroker
SOCRadar Research Team: Latest Activities of IntelBrokerSOCRadar Research Team: Latest Activities of IntelBroker
SOCRadar Research Team: Latest Activities of IntelBroker
SOCRadar
 
Enhancing Research Orchestration Capabilities at ORNL.pdf
Enhancing Research Orchestration Capabilities at ORNL.pdfEnhancing Research Orchestration Capabilities at ORNL.pdf
Enhancing Research Orchestration Capabilities at ORNL.pdf
Globus
 
AI Pilot Review: The World’s First Virtual Assistant Marketing Suite
AI Pilot Review: The World’s First Virtual Assistant Marketing SuiteAI Pilot Review: The World’s First Virtual Assistant Marketing Suite
AI Pilot Review: The World’s First Virtual Assistant Marketing Suite
Google
 
Cyaniclab : Software Development Agency Portfolio.pdf
Cyaniclab : Software Development Agency Portfolio.pdfCyaniclab : Software Development Agency Portfolio.pdf
Cyaniclab : Software Development Agency Portfolio.pdf
Cyanic lab
 
2024 RoOUG Security model for the cloud.pptx
2024 RoOUG Security model for the cloud.pptx2024 RoOUG Security model for the cloud.pptx
2024 RoOUG Security model for the cloud.pptx
Georgi Kodinov
 
Orion Context Broker introduction 20240604
Orion Context Broker introduction 20240604Orion Context Broker introduction 20240604
Orion Context Broker introduction 20240604
Fermin Galan
 
Developing Distributed High-performance Computing Capabilities of an Open Sci...
Developing Distributed High-performance Computing Capabilities of an Open Sci...Developing Distributed High-performance Computing Capabilities of an Open Sci...
Developing Distributed High-performance Computing Capabilities of an Open Sci...
Globus
 
First Steps with Globus Compute Multi-User Endpoints
First Steps with Globus Compute Multi-User EndpointsFirst Steps with Globus Compute Multi-User Endpoints
First Steps with Globus Compute Multi-User Endpoints
Globus
 
Vitthal Shirke Microservices Resume Montevideo
Vitthal Shirke Microservices Resume MontevideoVitthal Shirke Microservices Resume Montevideo
Vitthal Shirke Microservices Resume Montevideo
Vitthal Shirke
 

Recently uploaded (20)

Innovating Inference - Remote Triggering of Large Language Models on HPC Clus...
Innovating Inference - Remote Triggering of Large Language Models on HPC Clus...Innovating Inference - Remote Triggering of Large Language Models on HPC Clus...
Innovating Inference - Remote Triggering of Large Language Models on HPC Clus...
 
Enterprise Resource Planning System in Telangana
Enterprise Resource Planning System in TelanganaEnterprise Resource Planning System in Telangana
Enterprise Resource Planning System in Telangana
 
Beyond Event Sourcing - Embracing CRUD for Wix Platform - Java.IL
Beyond Event Sourcing - Embracing CRUD for Wix Platform - Java.ILBeyond Event Sourcing - Embracing CRUD for Wix Platform - Java.IL
Beyond Event Sourcing - Embracing CRUD for Wix Platform - Java.IL
 
Understanding Globus Data Transfers with NetSage
Understanding Globus Data Transfers with NetSageUnderstanding Globus Data Transfers with NetSage
Understanding Globus Data Transfers with NetSage
 
Globus Connect Server Deep Dive - GlobusWorld 2024
Globus Connect Server Deep Dive - GlobusWorld 2024Globus Connect Server Deep Dive - GlobusWorld 2024
Globus Connect Server Deep Dive - GlobusWorld 2024
 
Webinar: Salesforce Document Management 2.0 - Smarter, Faster, Better
Webinar: Salesforce Document Management 2.0 - Smarter, Faster, BetterWebinar: Salesforce Document Management 2.0 - Smarter, Faster, Better
Webinar: Salesforce Document Management 2.0 - Smarter, Faster, Better
 
Navigating the Metaverse: A Journey into Virtual Evolution"
Navigating the Metaverse: A Journey into Virtual Evolution"Navigating the Metaverse: A Journey into Virtual Evolution"
Navigating the Metaverse: A Journey into Virtual Evolution"
 
Corporate Management | Session 3 of 3 | Tendenci AMS
Corporate Management | Session 3 of 3 | Tendenci AMSCorporate Management | Session 3 of 3 | Tendenci AMS
Corporate Management | Session 3 of 3 | Tendenci AMS
 
Accelerate Enterprise Software Engineering with Platformless
Accelerate Enterprise Software Engineering with PlatformlessAccelerate Enterprise Software Engineering with Platformless
Accelerate Enterprise Software Engineering with Platformless
 
Using IESVE for Room Loads Analysis - Australia & New Zealand
Using IESVE for Room Loads Analysis - Australia & New ZealandUsing IESVE for Room Loads Analysis - Australia & New Zealand
Using IESVE for Room Loads Analysis - Australia & New Zealand
 
Climate Science Flows: Enabling Petabyte-Scale Climate Analysis with the Eart...
Climate Science Flows: Enabling Petabyte-Scale Climate Analysis with the Eart...Climate Science Flows: Enabling Petabyte-Scale Climate Analysis with the Eart...
Climate Science Flows: Enabling Petabyte-Scale Climate Analysis with the Eart...
 
SOCRadar Research Team: Latest Activities of IntelBroker
SOCRadar Research Team: Latest Activities of IntelBrokerSOCRadar Research Team: Latest Activities of IntelBroker
SOCRadar Research Team: Latest Activities of IntelBroker
 
Enhancing Research Orchestration Capabilities at ORNL.pdf
Enhancing Research Orchestration Capabilities at ORNL.pdfEnhancing Research Orchestration Capabilities at ORNL.pdf
Enhancing Research Orchestration Capabilities at ORNL.pdf
 
AI Pilot Review: The World’s First Virtual Assistant Marketing Suite
AI Pilot Review: The World’s First Virtual Assistant Marketing SuiteAI Pilot Review: The World’s First Virtual Assistant Marketing Suite
AI Pilot Review: The World’s First Virtual Assistant Marketing Suite
 
Cyaniclab : Software Development Agency Portfolio.pdf
Cyaniclab : Software Development Agency Portfolio.pdfCyaniclab : Software Development Agency Portfolio.pdf
Cyaniclab : Software Development Agency Portfolio.pdf
 
2024 RoOUG Security model for the cloud.pptx
2024 RoOUG Security model for the cloud.pptx2024 RoOUG Security model for the cloud.pptx
2024 RoOUG Security model for the cloud.pptx
 
Orion Context Broker introduction 20240604
Orion Context Broker introduction 20240604Orion Context Broker introduction 20240604
Orion Context Broker introduction 20240604
 
Developing Distributed High-performance Computing Capabilities of an Open Sci...
Developing Distributed High-performance Computing Capabilities of an Open Sci...Developing Distributed High-performance Computing Capabilities of an Open Sci...
Developing Distributed High-performance Computing Capabilities of an Open Sci...
 
First Steps with Globus Compute Multi-User Endpoints
First Steps with Globus Compute Multi-User EndpointsFirst Steps with Globus Compute Multi-User Endpoints
First Steps with Globus Compute Multi-User Endpoints
 
Vitthal Shirke Microservices Resume Montevideo
Vitthal Shirke Microservices Resume MontevideoVitthal Shirke Microservices Resume Montevideo
Vitthal Shirke Microservices Resume Montevideo
 

Project_Automatic Photo Classification Web Service

  • 1. 17-03-31 SookmyungGo Girlclash 1 Girlclash SookmyungGo Girlclash Automatic Photo Classification Web Service 숙명고팀 | 방누리 신아영 이효정 지도교수 | 최영우 교수님 G
  • 2. 17-03-31 SookmyungGo Girlclash 2 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
  • 3. 17-03-31 SookmyungGo Girlclash 3 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
  • 4. 17-03-31 SookmyungGo Girlclash 4 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
  • 5. 17-03-31 SookmyungGo Girlclash 5 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
  • 6. 17-03-31 SookmyungGo Girlclash 6 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
  • 7. 17-03-31 SookmyungGo Girlclash 7 What is Girlclash?
  • 8. 17-03-31 SookmyungGo Girlclash 8 What is Girlclash? Deep Learning SMGO
  • 9. 17-03-31 SookmyungGo Girlclash 9 What is Girlclash?
  • 10. 17-03-31 SookmyungGo Girlclash 10 What is Girlclash? Automatic Photo Classification Web Service
  • 11. 17-03-31 SookmyungGo Girlclash 11 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
  • 12. 17-03-31 SookmyungGo Girlclash 12 How does it work? – Tools DJANGO TENSORFLOW CNN MODEL Python Web Google’s Machine Learning Library Convolutional Neural Network for Image Process
  • 13. 17-03-31 SookmyungGo Girlclash 13 How does it work? – Tools DJANGO Python Web OBJECT RELATIONAL MAPPING •다양한 DB 지원 •유연성 •다양한 내장 함수 BACKEND CONSOLE •콘솔 상 DB 관리 가능 •NoSQL MODULE •프로젝트-모듈 분리 •전체 개발시간 단축
  • 14. 17-03-31 SookmyungGo Girlclash 14 How does it work? – Tools TENSORFLOW Google’s Machine Learning Library STRENGTH •Tensorboard로 구조도 제공 •사용자 정의 신경망 가능 •복잡한 연산 자동 수행 DEFINITION •Tensor란? 동적 다차원 데이터 배열 •노드 = 연산 엣지 = 데이터 전달 데이터 = Tensor GRAPH OPERATION
  • 15. 17-03-31 SookmyungGo Girlclash 15 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process LENET 1998 ALEXNET 2012 •Handwritten digits (Grayscale) •Average pooling Layer •Sigmoid function •RGB images •Max pooling layer •ReLU function
  • 16. 17-03-31 SookmyungGo Girlclash 16 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process LENET 1998 ALEXNET 2012 •Handwritten digits (Grayscale) •Average pooling Layer •Sigmoid function •RGB images •Max pooling layer •ReLU function
  • 17. 17-03-31 SookmyungGo Girlclash 17 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •Dropout A B C
  • 18. 17-03-31 SookmyungGo Girlclash 18 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •Dropout
  • 19. 17-03-31 SookmyungGo Girlclash 19 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012
  • 20. 17-03-31 SookmyungGo Girlclash 20 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •Dropout
  • 21. 17-03-31 SookmyungGo Girlclash 21 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •Dropout 50 100 Sigmoid ReLU 정확도(%) 활성화함수 활성화 함수에 따른 정확도(%)
  • 22. 17-03-31 SookmyungGo Girlclash 22 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •Dropout
  • 23. 17-03-31 SookmyungGo Girlclash 23 How does it work? – Tools CNN MODEL Convolutional Neural Network for Image Process ALEXNET 2012 •RGB images •Maxpooling layer •ReLU function •LRN & Dropout
  • 24. 17-03-31 SookmyungGo Girlclash 24 참고. CNN 학습 모델의 형태 (tensorboard)
  • 25. 17-03-31 SookmyungGo Girlclash 25 참고. CNN 학습 모델의 형태 (diagram)
  • 26. 17-03-31 SookmyungGo Girlclash 26 How does it work? – Machine Learning Environment Learning Conditions CNNs Misclassification CPU GPU LeNet AlexNet Time Rate Dataset Solution
  • 27. 17-03-31 SookmyungGo Girlclash 27 How does it work? – Machine Learning ENVIRONMENT 60 3 CPU GPU 시간(분) 환경 학습 환경에 따른 소요 시간(분) CPU GPU • 데이터셋 : 약 40,000장 • 신경망 : LeNet • 학습 횟수 : 8,500회 GTX1070 with
  • 28. 17-03-31 SookmyungGo Girlclash 28 How does it work? – Machine Learning • 5-class Dataset Faces 9,958 Pets 12,474 Food 9,866 Nature 8,703 Fashion 11,786 LEARNING CONDITION - DATASET Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Dataset Rate Time
  • 29. 17-03-31 SookmyungGo Girlclash 29 How does it work? – Machine Learning • Learning Rate LEARNING CONDITION – LEARNING RATE Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Dataset Rate Time45.5 41.5 52.8 45.8 46.8 0.001 0.003 0.005 0.007 0.01 정확도(%) 학습율 학습율에 따른 정확도(%) • 데이터셋 : 약 40,000장 • 신경망 : AlexNet • 학습 횟수 : 5,000회 ~
  • 30. 17-03-31 SookmyungGo Girlclash 30 How does it work? – Machine Learning • Learning Time LEARNING CONDITION – LEARNING TIME Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Dataset Rate Time71.8 93.7 93.7 85.9 96.3 97.8 8,500회 100,000회 500,000회 정확도(%) 학습횟수 학습 횟수에 따른 정확도(%) train test • 데이터셋 : 약 40,000장 • 신경망 : Alexnet
  • 31. 17-03-31 SookmyungGo Girlclash 31 How does it work? – Machine Learning • LeNet • AlexNet CNNS Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum 60 3 CPU GPU 정확도(%) 신경망 신경망에 따른 정확도(%) 열1 Series 1 • 데이터셋 : 약 40,000장 • 학습 횟수 : 8,500회 66.8 71.8 72 85.9 LeNet AlexNet 정확도(%) 신경망 신경망에 따른 정확도(%) train test
  • 32. 17-03-31 SookmyungGo Girlclash 32 How does it work? – Machine Learning 1. 오분류 해결 방법 A. Prediction : 예측치 < 0.5인 이미지를 Etc로 분류 B. Garbage : Garbage Class를 포함하여 학습 MISCLASSIFICATION Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Solution Test case Soccer ball Umbrella Airplane Car 개수 63 75 800 8144 Garbage class 내 포함 여부 미포함 포함 미포함이지만 유사함
  • 33. 17-03-31 SookmyungGo Girlclash 33 How does it work? – Machine Learning 1. 오분류 해결 방법 A. Prediction : 예측치 < 0.5인 이미지를 Etc로 분류 B. Garbage : Garbage Class를 포함하여 학습 MISCLASSIFICATION Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Solution 1 2 3 4 5 6 (garbage)
  • 34. 17-03-31 SookmyungGo Girlclash 34 How does it work? – Machine Learning 1. 오분류 해결 방법 A. Prediction : 예측치 < 0.5인 이미지를 Etc로 분류 B. Garbage : Garbage Class를 포함하여 학습 2. Test case MISCLASSIFICATION Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Solution Test case Soccer ball Umbrella Airplane Car 개수 63 75 800 8144 Garbage class 내 포함 여부 미포함 포함 미포함이지만 유사함
  • 35. 17-03-31 SookmyungGo Girlclash 35 How does it work? – Machine Learning MISCLASSIFICATION Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Solution 31.3 35.9 44 42.7 94.9 1.5 36.3 35.5 Garbage Prediction 성공률(%) 오분류해결방법 해결 방법에 따른 분류 성공률(%) Car Airplane Umbrella Soccer ball • 학습 횟수 : 100,000회 • 신경망 : AlexnetA B
  • 36. 17-03-31 SookmyungGo Girlclash 36 How does it work? – Machine Learning MISCLASSIFICATION Lorem ipsum Lorem ipsum Lorem Ipsum Lorem Ipsum Solution 1 2 3 4 5 6 (garbage) B
  • 37. 17-03-31 SookmyungGo Girlclash 37 How does it work? – Web Application 시연 영상 및 설명
  • 38. 17-03-31 SookmyungGo Girlclash 38 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
  • 39. 17-03-31 SookmyungGo Girlclash 39 Things to improve 압축 파일 다운로드 기능 추가 오분류 개선 회원 관리 기능 앱 개발
  • 40. 17-03-31 SookmyungGo Girlclash 40 Index What is Girlclash? How does it work? Things to improve What we expect 1 2 3 4
  • 41. 17-03-31 SookmyungGo Girlclash 41 What we expect SNS로 확장 사진 추천 기능 추가
  • 42. 17-03-31 SookmyungGo Girlclash 42 Thank you for your attention! Do you have any questions? 방누리 신아영 이효정 웹 프론트엔드 /백엔드 데이터셋 수집 /웹 프론트엔드 딥러닝 /웹과 분류코드 연결
  • 43. 17-03-31 SookmyungGo Girlclash 43 • CNN 학습 모델 - http://dsmoon.tistory.com/entry/TensorFlow-tutorial-Convolutional-Neural-Networks • LeNet과 AlexNet - 밑바닥부터 시작하는 딥러닝 (한빛미디어, 사이토 고키, 2017) • CNN - http://hamait.tistory.com/535 • 활성화 함수 - 모두의 딥러닝 (김성훈) • 드롭아웃 - http://bcho.tistory.com/tag/alexnet • 풀링 - https://tensorflow.blog/%ED%95%B4%EC%BB%A4%EC%97%90%EA%B2%8C- %EC%A0%84%ED%95%B4%EB%93%A4%EC%9D%80- %EB%A8%B8%EC%8B%A0%EB%9F%AC%EB%8B%9D-4/ - https://www.embedded-vision.com/platinum-members/cadence/embedded-vision- training/documents/pages/neuralnetworksimagerecognition - http://sanghyukchun.github.io/75/ 출처