SlideShare a Scribd company logo
Computer Vision
with Deep Learning
OOP Munich, 07.02.2018, Andreas Eßbaumer
Computer Vision with Deep Learning | Andreas Eßbaumer
▪ History and Hype of Deep Learning and Convolution Neural Networks
▪ Visualization of Convolutions
▪ The three and a half types of Computer Vision applications
▪ How to implement Deep Learning for Computer Vision?
▪ Image Classification: Pipelines
▪ Object Detection: Filler caps
© 2017 Capgemini. All rights reserved. 2
Overview
Computer Vision with Deep Learning | Andreas Eßbaumer 3© 2018 Capgemini. All rights reserved.
A bit of
history
Source: Bennamoun, DeepLearn 2017
Computer Vision with Deep Learning | Andreas Eßbaumer 4© 2018 Capgemini. All rights reserved.
A bit of
history
Source: Bennamoun, DeepLearn 2017
Computer Vision with Deep Learning | Andreas Eßbaumer
Why the hype? – The
ImageNet competition
ImageNet:
Over 10 Mio labeled images.
Over 10.000 different labels.
ImageNet Competition:
Restrict the number of labels to 1000.
Algorithms have to produce up to five labels per
image.
Computer Vision with Deep Learning | Andreas Eßbaumer 6© 2017 Capgemini. All rights reserved.
ILSVRC challenge
Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 7
Using convolutions the network is able to learn valuable filters and
extract the most important features
▪ Each convolutional layer generates
many different representations of
the image through filters
▪ Filters are in most cases of the sizes
3x3, 5x5 or 7x7 and are learned(!)
▪ Max-pooling areas reduces the size of
the image and makes sure filters cover
a larger area of the picture
▪ Generally the latter the layer the
larger the area of an image which is
covered by a filter and the higher the
number of filters
10
Layer 1 + 2
3 4
5 6 7
8 9
13
14 15 16
1211
The neural net
architecture of VGG 16
Image Source: https://blog.heuritech.com/2016/02/29/a-brief-report-of-
the-heuritech-deep-learning-meetup-5/
Computer Vision with Deep Learning | Andreas Eßbaumer
Visualization of convolution layers learned(!) by the network
From millions of training images, the network learns that the most basic features to
look for in an image are basic forms like edges, circles or smooth color transitions.
Source: Visualizing and Understanding Convolutional Networks by Zeiler/Fergus arXiv:1311.2901
Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 9
Visualization of convolution layers learned(!) by the network
• Each convolutional layer
consists of a high number
of filters
• Depending on the image
some filters are more
and differently
applicable than others
Source: https://github.com/yosinski/deep-visualization-toolbox
https://www.youtube.com/watch?v=AgkfIQ4IGaM
Computer Vision with Deep Learning | Andreas Eßbaumer
Visualization of convolution layers learned(!) by the network
Working with the most basic features extracted by the first layers, the network is able
to combine them in a larger area of the picture and find more complex
structures like certain patterns and written text pieces.
Source: Visualizing and Understanding Convolutional Networks by Zeiler/Fergus arXiv:1311.2901
Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 11
Visualization of convolution layers learned(!) by the network
the number of
filters increases
to hundreds
of specialized
filters
Source: Visualizing and Understanding Convolutional Networks by Zeiler/Fergus arXiv:1311.2901
Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 12
Visualization of convolution layers learned(!) by the network
• The filters are
purely learned via
training data.
• There is even a filter
highly specialized
on detection
human faces.
Source: https://github.com/yosinski/deep-visualization-toolbox
https://www.youtube.com/watch?v=AgkfIQ4IGaM
Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 13
There are three and a half types of computer vision applications
Image Classification/Recognition Object Detection
Image Transformation/FusionDetect similar images
Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 14
The two major application types are
image classification and object detection
Image Classification/Recognition Object Detection
• Evaluates the image as a whole
• Can categorize different states or
even abstract concepts
• Looks for physical or at least visually
existing objects
• Returns multiple objects of various
types and with their position
Computer Vision with Deep Learning | Andreas Eßbaumer
<
15© 2017 Capgemini. All rights reserved.Source: fast.ai
• state-of-the-art image
classification model in 7
lines of code.
• prerequisites are sufficient,
representative, labelled
training images and GPUs
• high-level deep learning
framework like Keras
make CNNs easy to run
• DL-Frameworks like
run CNNs efficiently on
GPUs.
Implementing an Image Classification model is easy
(at least easier than it seems)
Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 16
To build an object detection model one can use the
Tensorflow Object Detection API
• No changes on the deep learning
code is necessary
• the training runs efficiently on
GPUs because of Tensorflow
• labelling by giving meta information
about which types of objects can be
found and where (in which box).
• Labelling tools are available, e.g.
labelImg
• Leveraging pre-trained models
Useful links: https://github.com/tensorflow/models/tree/master/research/object_detection
https://github.com/tzutalin/labelImg
Computer Vision with Deep Learning | Andreas Eßbaumer 17© 2017 Capgemini. All rights reserved.
• 3.5 million km of pipelines in the world
(2014)
• Many of these are more than 50 years old.
• Pipelines operate under immense pressure:
Even small cracks can have disastrous
consequences
An example: Defect classification for pipeline maintenance
• Our client offers solutions for pipeline
maintenance
• According to regulation pipeline inspections
need to be done every few years
• pipeline inspection robots for non-
destructive testing of pipeline integrity are
used
Computer Vision with Deep Learning | Andreas Eßbaumer 18© 2017 Capgemini. All rights reserved.
• Defect detection is very challenging
• Firstly, the sensors produce huge amounts
of data that need to be preprocessed and
stored online, i.e. during the inspection run.
• Secondly, analyzing the images requires
special training and is very time intensive.
Defect classification with ultrasound
• Given an image of a region of
interest, can you accurately predict
the defect type?
The use case
A pipeline inspection robot in its rare clean state
Computer Vision with Deep Learning | Andreas Eßbaumer
Deep Learning gave better results with little effort
0,00%
10,00%
20,00%
30,00%
40,00%
50,00%
60,00%
70,00%
80,00%
90,00%
100,00%
CV InceptionV3 - 172+ InceptionV3 - 0+ Custom-V1
Defect classification results
Recall Specificity
• After the successful POC, industrialization is currently ongoing
• Human experts have a strong assistance tool now
• Lowering risks by double-checks (Human + AI)
Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 20
We use object detection to identify different types of defects
and their position on a filler cap
• Using a deep learning
pipeline of two models
to reach full flexibility
• The first object detection
model runs
continuously and looks
for various objects e.g.
filler caps.
• a second object
detection model looks
for various defects on
the surface of the filler
cap
Model 1 Model 2
Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 21
Depending on the number and types of defects the quality
check is passed or failed
• invariant to different
environmental lighting,
backgrounds, angles,
reflections etc.
• Detects small defects like
scratches, too much paint or
punctual damages.
• Even the positions of the defects
are given by the model providing.
ideal information for potential
rework
• Built in one day
• Uses 100 labelled images per
filler cap
Computer Vision with Deep Learning | Andreas Eßbaumer
Andreas Eßbaumer
Senior Data Scientist
Capgemini Munich
Olof-Palme-Straße 14
81829 Munich
andreas.essbaumer@capgemini.com
M: +49 151 4025 2219
Summary
Convolutional Neural Networks (CNNs)
• are the State-of-the-Art for Computer Vision
• run via Open-Source-Frameworks
• require GPUs and labelled Images
• are (relatively) easy to run
A global leader in consulting, technology services and digital transformation,
Capgemini is at the forefront of innovation to address the entire breadth of clients’
opportunities in the evolving world of cloud, digital and platforms. Building on its
strong 50-year heritage and deep industry-specific expertise, Capgemini enables
organizations to realize their business ambitions through an array of services from
strategy to operations. Capgemini is driven by the conviction that the business
value of technology comes from and through people. It is a multicultural company
of 200,000 team members in over 40 countries. The Group reported 2016 global
revenues of EUR 12.5 billion.
About Capgemini
Learn more about us at
www.capgemini.com
This message contains information that may be privileged or confidential and is
the property of the Capgemini Group.
Copyright © 2017 Capgemini. All rights reserved.
People matter, results count.

More Related Content

What's hot

Convolutional Neural Network (CNN) presentation from theory to code in Theano
Convolutional Neural Network (CNN) presentation from theory to code in TheanoConvolutional Neural Network (CNN) presentation from theory to code in Theano
Convolutional Neural Network (CNN) presentation from theory to code in TheanoSeongwon Hwang
 
Image Processing with OpenCV
Image Processing with OpenCVImage Processing with OpenCV
Image Processing with OpenCVdebayanin
 
Introduction to Deep Learning
Introduction to Deep LearningIntroduction to Deep Learning
Introduction to Deep LearningOswald Campesato
 
Deep Belief Networks
Deep Belief NetworksDeep Belief Networks
Deep Belief NetworksHasan H Topcu
 
Deep learning - A Visual Introduction
Deep learning - A Visual IntroductionDeep learning - A Visual Introduction
Deep learning - A Visual IntroductionLukas Masuch
 
Transfer Learning and Fine-tuning Deep Neural Networks
 Transfer Learning and Fine-tuning Deep Neural Networks Transfer Learning and Fine-tuning Deep Neural Networks
Transfer Learning and Fine-tuning Deep Neural NetworksPyData
 
Human age and gender Detection
Human age and gender DetectionHuman age and gender Detection
Human age and gender DetectionAbhiAchalla
 
Image classification using convolutional neural network
Image classification using convolutional neural networkImage classification using convolutional neural network
Image classification using convolutional neural networkKIRAN R
 
Computer vision, machine, and deep learning
Computer vision, machine, and deep learningComputer vision, machine, and deep learning
Computer vision, machine, and deep learningIgi Ardiyanto
 
Deep Learning: Application Landscape - March 2018
Deep Learning: Application Landscape - March 2018Deep Learning: Application Landscape - March 2018
Deep Learning: Application Landscape - March 2018Grigory Sapunov
 
Facial Expression Recognition System using Deep Convolutional Neural Networks.
Facial Expression Recognition  System using Deep Convolutional Neural Networks.Facial Expression Recognition  System using Deep Convolutional Neural Networks.
Facial Expression Recognition System using Deep Convolutional Neural Networks.Sandeep Wakchaure
 
Image Segmentation Using Deep Learning : A survey
Image Segmentation Using Deep Learning : A surveyImage Segmentation Using Deep Learning : A survey
Image Segmentation Using Deep Learning : A surveyNUPUR YADAV
 
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)Fellowship at Vodafone FutureLab
 
Deep Learning - Overview of my work II
Deep Learning - Overview of my work IIDeep Learning - Overview of my work II
Deep Learning - Overview of my work IIMohamed Loey
 
Convolutional neural network
Convolutional neural networkConvolutional neural network
Convolutional neural networkItachi SK
 

What's hot (20)

Convolutional Neural Network (CNN) presentation from theory to code in Theano
Convolutional Neural Network (CNN) presentation from theory to code in TheanoConvolutional Neural Network (CNN) presentation from theory to code in Theano
Convolutional Neural Network (CNN) presentation from theory to code in Theano
 
Image Processing with OpenCV
Image Processing with OpenCVImage Processing with OpenCV
Image Processing with OpenCV
 
Opencv
OpencvOpencv
Opencv
 
Introduction to Deep Learning
Introduction to Deep LearningIntroduction to Deep Learning
Introduction to Deep Learning
 
Deep Belief Networks
Deep Belief NetworksDeep Belief Networks
Deep Belief Networks
 
Deep learning - A Visual Introduction
Deep learning - A Visual IntroductionDeep learning - A Visual Introduction
Deep learning - A Visual Introduction
 
Transfer Learning and Fine-tuning Deep Neural Networks
 Transfer Learning and Fine-tuning Deep Neural Networks Transfer Learning and Fine-tuning Deep Neural Networks
Transfer Learning and Fine-tuning Deep Neural Networks
 
Human age and gender Detection
Human age and gender DetectionHuman age and gender Detection
Human age and gender Detection
 
Image classification using convolutional neural network
Image classification using convolutional neural networkImage classification using convolutional neural network
Image classification using convolutional neural network
 
Computer vision, machine, and deep learning
Computer vision, machine, and deep learningComputer vision, machine, and deep learning
Computer vision, machine, and deep learning
 
Transfer Learning
Transfer LearningTransfer Learning
Transfer Learning
 
Deep Learning: Application Landscape - March 2018
Deep Learning: Application Landscape - March 2018Deep Learning: Application Landscape - March 2018
Deep Learning: Application Landscape - March 2018
 
Facial Expression Recognition System using Deep Convolutional Neural Networks.
Facial Expression Recognition  System using Deep Convolutional Neural Networks.Facial Expression Recognition  System using Deep Convolutional Neural Networks.
Facial Expression Recognition System using Deep Convolutional Neural Networks.
 
Image Segmentation Using Deep Learning : A survey
Image Segmentation Using Deep Learning : A surveyImage Segmentation Using Deep Learning : A survey
Image Segmentation Using Deep Learning : A survey
 
Cnn
CnnCnn
Cnn
 
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
 
Deep Learning - Overview of my work II
Deep Learning - Overview of my work IIDeep Learning - Overview of my work II
Deep Learning - Overview of my work II
 
Introduction to OpenCV
Introduction to OpenCVIntroduction to OpenCV
Introduction to OpenCV
 
Deep learning presentation
Deep learning presentationDeep learning presentation
Deep learning presentation
 
Convolutional neural network
Convolutional neural networkConvolutional neural network
Convolutional neural network
 

Similar to Computer Vision with Deep Learning

Makine Öğrenmesi ile Görüntü Tanıma | Image Recognition using Machine Learning
Makine Öğrenmesi ile Görüntü Tanıma | Image Recognition using Machine LearningMakine Öğrenmesi ile Görüntü Tanıma | Image Recognition using Machine Learning
Makine Öğrenmesi ile Görüntü Tanıma | Image Recognition using Machine LearningAli Alkan
 
CAR DAMAGE DETECTION USING DEEP LEARNING
CAR DAMAGE DETECTION USING DEEP LEARNINGCAR DAMAGE DETECTION USING DEEP LEARNING
CAR DAMAGE DETECTION USING DEEP LEARNINGIRJET Journal
 
Color Detection & Segmentation based Invisible Cloak
Color Detection & Segmentation based Invisible CloakColor Detection & Segmentation based Invisible Cloak
Color Detection & Segmentation based Invisible CloakAviral Chaurasia
 
IEEE EED2021 AI use cases in Computer Vision
IEEE EED2021 AI use cases in Computer VisionIEEE EED2021 AI use cases in Computer Vision
IEEE EED2021 AI use cases in Computer VisionSAMeh Zaghloul
 
Machine Learning Model for Gender Detection
Machine Learning Model for Gender DetectionMachine Learning Model for Gender Detection
Machine Learning Model for Gender DetectionTecnoIncentive
 
“Tools and Strategies for Quickly Building Effective Image Datasets,” a Prese...
“Tools and Strategies for Quickly Building Effective Image Datasets,” a Prese...“Tools and Strategies for Quickly Building Effective Image Datasets,” a Prese...
“Tools and Strategies for Quickly Building Effective Image Datasets,” a Prese...Edge AI and Vision Alliance
 
Computer architecture for vision system
Computer architecture for vision systemComputer architecture for vision system
Computer architecture for vision systemAkashPatil334
 
The Power of Scripting with a Desktop SEM
The Power of Scripting with a Desktop SEM The Power of Scripting with a Desktop SEM
The Power of Scripting with a Desktop SEM InsideScientific
 
Image Processing In Open CV. Image Processing In Open CV. Image Processing In...
Image Processing In Open CV. Image Processing In Open CV. Image Processing In...Image Processing In Open CV. Image Processing In Open CV. Image Processing In...
Image Processing In Open CV. Image Processing In Open CV. Image Processing In...Antoinette Williams
 
AUTOMATIC APPEARANCE MASK AND BODY TEMPERATURE FINDING SYSTEM
AUTOMATIC APPEARANCE MASK AND BODY TEMPERATURE FINDING SYSTEMAUTOMATIC APPEARANCE MASK AND BODY TEMPERATURE FINDING SYSTEM
AUTOMATIC APPEARANCE MASK AND BODY TEMPERATURE FINDING SYSTEMIRJET Journal
 
FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDS
FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDSFACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDS
FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDSIRJET Journal
 
2 D3 D Concersion Swaggmedia
2 D3 D Concersion   Swaggmedia2 D3 D Concersion   Swaggmedia
2 D3 D Concersion SwaggmediaCraig Nobles
 
A DEEP LEARNING APPROACH FOR SEMANTIC SEGMENTATION IN BRAIN TUMOR IMAGES
A DEEP LEARNING APPROACH FOR SEMANTIC SEGMENTATION IN BRAIN TUMOR IMAGESA DEEP LEARNING APPROACH FOR SEMANTIC SEGMENTATION IN BRAIN TUMOR IMAGES
A DEEP LEARNING APPROACH FOR SEMANTIC SEGMENTATION IN BRAIN TUMOR IMAGESPNandaSai
 
IRJET- Homomorphic Image Encryption
IRJET-  	  Homomorphic Image EncryptionIRJET-  	  Homomorphic Image Encryption
IRJET- Homomorphic Image EncryptionIRJET Journal
 
Motion detection system
Motion detection systemMotion detection system
Motion detection systemWritingHubUK
 
"Deep Learning Beyond Cats and Cars: Developing a Real-life DNN-based Embedde...
"Deep Learning Beyond Cats and Cars: Developing a Real-life DNN-based Embedde..."Deep Learning Beyond Cats and Cars: Developing a Real-life DNN-based Embedde...
"Deep Learning Beyond Cats and Cars: Developing a Real-life DNN-based Embedde...Edge AI and Vision Alliance
 
Efficient video perception through AI
Efficient video perception through AIEfficient video perception through AI
Efficient video perception through AIQualcomm Research
 
Smart Face Recognition System Analysis
Smart Face Recognition System AnalysisSmart Face Recognition System Analysis
Smart Face Recognition System AnalysisVishal Aditya
 

Similar to Computer Vision with Deep Learning (20)

Makine Öğrenmesi ile Görüntü Tanıma | Image Recognition using Machine Learning
Makine Öğrenmesi ile Görüntü Tanıma | Image Recognition using Machine LearningMakine Öğrenmesi ile Görüntü Tanıma | Image Recognition using Machine Learning
Makine Öğrenmesi ile Görüntü Tanıma | Image Recognition using Machine Learning
 
OpenPOWER/POWER9 AI webinar
OpenPOWER/POWER9 AI webinar OpenPOWER/POWER9 AI webinar
OpenPOWER/POWER9 AI webinar
 
CAR DAMAGE DETECTION USING DEEP LEARNING
CAR DAMAGE DETECTION USING DEEP LEARNINGCAR DAMAGE DETECTION USING DEEP LEARNING
CAR DAMAGE DETECTION USING DEEP LEARNING
 
Color Detection & Segmentation based Invisible Cloak
Color Detection & Segmentation based Invisible CloakColor Detection & Segmentation based Invisible Cloak
Color Detection & Segmentation based Invisible Cloak
 
IEEE EED2021 AI use cases in Computer Vision
IEEE EED2021 AI use cases in Computer VisionIEEE EED2021 AI use cases in Computer Vision
IEEE EED2021 AI use cases in Computer Vision
 
Machine Learning Model for Gender Detection
Machine Learning Model for Gender DetectionMachine Learning Model for Gender Detection
Machine Learning Model for Gender Detection
 
“Tools and Strategies for Quickly Building Effective Image Datasets,” a Prese...
“Tools and Strategies for Quickly Building Effective Image Datasets,” a Prese...“Tools and Strategies for Quickly Building Effective Image Datasets,” a Prese...
“Tools and Strategies for Quickly Building Effective Image Datasets,” a Prese...
 
Computer architecture for vision system
Computer architecture for vision systemComputer architecture for vision system
Computer architecture for vision system
 
Ijetr011814
Ijetr011814Ijetr011814
Ijetr011814
 
The Power of Scripting with a Desktop SEM
The Power of Scripting with a Desktop SEM The Power of Scripting with a Desktop SEM
The Power of Scripting with a Desktop SEM
 
Image Processing In Open CV. Image Processing In Open CV. Image Processing In...
Image Processing In Open CV. Image Processing In Open CV. Image Processing In...Image Processing In Open CV. Image Processing In Open CV. Image Processing In...
Image Processing In Open CV. Image Processing In Open CV. Image Processing In...
 
AUTOMATIC APPEARANCE MASK AND BODY TEMPERATURE FINDING SYSTEM
AUTOMATIC APPEARANCE MASK AND BODY TEMPERATURE FINDING SYSTEMAUTOMATIC APPEARANCE MASK AND BODY TEMPERATURE FINDING SYSTEM
AUTOMATIC APPEARANCE MASK AND BODY TEMPERATURE FINDING SYSTEM
 
FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDS
FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDSFACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDS
FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDS
 
2 D3 D Concersion Swaggmedia
2 D3 D Concersion   Swaggmedia2 D3 D Concersion   Swaggmedia
2 D3 D Concersion Swaggmedia
 
A DEEP LEARNING APPROACH FOR SEMANTIC SEGMENTATION IN BRAIN TUMOR IMAGES
A DEEP LEARNING APPROACH FOR SEMANTIC SEGMENTATION IN BRAIN TUMOR IMAGESA DEEP LEARNING APPROACH FOR SEMANTIC SEGMENTATION IN BRAIN TUMOR IMAGES
A DEEP LEARNING APPROACH FOR SEMANTIC SEGMENTATION IN BRAIN TUMOR IMAGES
 
IRJET- Homomorphic Image Encryption
IRJET-  	  Homomorphic Image EncryptionIRJET-  	  Homomorphic Image Encryption
IRJET- Homomorphic Image Encryption
 
Motion detection system
Motion detection systemMotion detection system
Motion detection system
 
"Deep Learning Beyond Cats and Cars: Developing a Real-life DNN-based Embedde...
"Deep Learning Beyond Cats and Cars: Developing a Real-life DNN-based Embedde..."Deep Learning Beyond Cats and Cars: Developing a Real-life DNN-based Embedde...
"Deep Learning Beyond Cats and Cars: Developing a Real-life DNN-based Embedde...
 
Efficient video perception through AI
Efficient video perception through AIEfficient video perception through AI
Efficient video perception through AI
 
Smart Face Recognition System Analysis
Smart Face Recognition System AnalysisSmart Face Recognition System Analysis
Smart Face Recognition System Analysis
 

More from Capgemini

Top Healthcare Trends 2022
Top Healthcare Trends 2022Top Healthcare Trends 2022
Top Healthcare Trends 2022Capgemini
 
Top P&C Insurance Trends 2022
Top P&C Insurance Trends 2022Top P&C Insurance Trends 2022
Top P&C Insurance Trends 2022Capgemini
 
Commercial Banking Trends book 2022
Commercial Banking Trends book 2022Commercial Banking Trends book 2022
Commercial Banking Trends book 2022Capgemini
 
Top Trends in Payments 2022
Top Trends in Payments 2022Top Trends in Payments 2022
Top Trends in Payments 2022Capgemini
 
Top Trends in Wealth Management 2022
Top Trends in Wealth Management 2022Top Trends in Wealth Management 2022
Top Trends in Wealth Management 2022Capgemini
 
Retail Banking Trends book 2022
Retail Banking Trends book 2022Retail Banking Trends book 2022
Retail Banking Trends book 2022Capgemini
 
Top Life Insurance Trends 2022
Top Life Insurance Trends 2022Top Life Insurance Trends 2022
Top Life Insurance Trends 2022Capgemini
 
キャップジェミニ、あなたの『RISE WITH SAP』のパートナーです
キャップジェミニ、あなたの『RISE WITH SAP』のパートナーですキャップジェミニ、あなたの『RISE WITH SAP』のパートナーです
キャップジェミニ、あなたの『RISE WITH SAP』のパートナーですCapgemini
 
Property & Casualty Insurance Top Trends 2021
Property & Casualty Insurance Top Trends 2021Property & Casualty Insurance Top Trends 2021
Property & Casualty Insurance Top Trends 2021Capgemini
 
Life Insurance Top Trends 2021
Life Insurance Top Trends 2021Life Insurance Top Trends 2021
Life Insurance Top Trends 2021Capgemini
 
Top Trends in Commercial Banking: 2021
Top Trends in Commercial Banking: 2021Top Trends in Commercial Banking: 2021
Top Trends in Commercial Banking: 2021Capgemini
 
Top Trends in Wealth Management: 2021
Top Trends in Wealth Management: 2021Top Trends in Wealth Management: 2021
Top Trends in Wealth Management: 2021Capgemini
 
Top Trends in Payments: 2021
Top Trends in Payments: 2021Top Trends in Payments: 2021
Top Trends in Payments: 2021Capgemini
 
Health Insurance Top Trends 2021
Health Insurance Top Trends 2021Health Insurance Top Trends 2021
Health Insurance Top Trends 2021Capgemini
 
Top Trends in Retail Banking: 2021
Top Trends in Retail Banking: 2021Top Trends in Retail Banking: 2021
Top Trends in Retail Banking: 2021Capgemini
 
Capgemini’s Connected Autonomous Planning
Capgemini’s Connected Autonomous PlanningCapgemini’s Connected Autonomous Planning
Capgemini’s Connected Autonomous PlanningCapgemini
 
Top Trends in Retail Banking: 2020
Top Trends in Retail Banking: 2020Top Trends in Retail Banking: 2020
Top Trends in Retail Banking: 2020Capgemini
 
Top Trends in Life Insurance: 2020
Top Trends in Life Insurance: 2020Top Trends in Life Insurance: 2020
Top Trends in Life Insurance: 2020Capgemini
 
Top Trends in Health Insurance: 2020
Top Trends in Health Insurance: 2020Top Trends in Health Insurance: 2020
Top Trends in Health Insurance: 2020Capgemini
 
Top Trends in Payments: 2020
Top Trends in Payments: 2020Top Trends in Payments: 2020
Top Trends in Payments: 2020Capgemini
 

More from Capgemini (20)

Top Healthcare Trends 2022
Top Healthcare Trends 2022Top Healthcare Trends 2022
Top Healthcare Trends 2022
 
Top P&C Insurance Trends 2022
Top P&C Insurance Trends 2022Top P&C Insurance Trends 2022
Top P&C Insurance Trends 2022
 
Commercial Banking Trends book 2022
Commercial Banking Trends book 2022Commercial Banking Trends book 2022
Commercial Banking Trends book 2022
 
Top Trends in Payments 2022
Top Trends in Payments 2022Top Trends in Payments 2022
Top Trends in Payments 2022
 
Top Trends in Wealth Management 2022
Top Trends in Wealth Management 2022Top Trends in Wealth Management 2022
Top Trends in Wealth Management 2022
 
Retail Banking Trends book 2022
Retail Banking Trends book 2022Retail Banking Trends book 2022
Retail Banking Trends book 2022
 
Top Life Insurance Trends 2022
Top Life Insurance Trends 2022Top Life Insurance Trends 2022
Top Life Insurance Trends 2022
 
キャップジェミニ、あなたの『RISE WITH SAP』のパートナーです
キャップジェミニ、あなたの『RISE WITH SAP』のパートナーですキャップジェミニ、あなたの『RISE WITH SAP』のパートナーです
キャップジェミニ、あなたの『RISE WITH SAP』のパートナーです
 
Property & Casualty Insurance Top Trends 2021
Property & Casualty Insurance Top Trends 2021Property & Casualty Insurance Top Trends 2021
Property & Casualty Insurance Top Trends 2021
 
Life Insurance Top Trends 2021
Life Insurance Top Trends 2021Life Insurance Top Trends 2021
Life Insurance Top Trends 2021
 
Top Trends in Commercial Banking: 2021
Top Trends in Commercial Banking: 2021Top Trends in Commercial Banking: 2021
Top Trends in Commercial Banking: 2021
 
Top Trends in Wealth Management: 2021
Top Trends in Wealth Management: 2021Top Trends in Wealth Management: 2021
Top Trends in Wealth Management: 2021
 
Top Trends in Payments: 2021
Top Trends in Payments: 2021Top Trends in Payments: 2021
Top Trends in Payments: 2021
 
Health Insurance Top Trends 2021
Health Insurance Top Trends 2021Health Insurance Top Trends 2021
Health Insurance Top Trends 2021
 
Top Trends in Retail Banking: 2021
Top Trends in Retail Banking: 2021Top Trends in Retail Banking: 2021
Top Trends in Retail Banking: 2021
 
Capgemini’s Connected Autonomous Planning
Capgemini’s Connected Autonomous PlanningCapgemini’s Connected Autonomous Planning
Capgemini’s Connected Autonomous Planning
 
Top Trends in Retail Banking: 2020
Top Trends in Retail Banking: 2020Top Trends in Retail Banking: 2020
Top Trends in Retail Banking: 2020
 
Top Trends in Life Insurance: 2020
Top Trends in Life Insurance: 2020Top Trends in Life Insurance: 2020
Top Trends in Life Insurance: 2020
 
Top Trends in Health Insurance: 2020
Top Trends in Health Insurance: 2020Top Trends in Health Insurance: 2020
Top Trends in Health Insurance: 2020
 
Top Trends in Payments: 2020
Top Trends in Payments: 2020Top Trends in Payments: 2020
Top Trends in Payments: 2020
 

Recently uploaded

IoT Analytics Company Presentation May 2024
IoT Analytics Company Presentation May 2024IoT Analytics Company Presentation May 2024
IoT Analytics Company Presentation May 2024IoTAnalytics
 
UiPath Test Automation using UiPath Test Suite series, part 1
UiPath Test Automation using UiPath Test Suite series, part 1UiPath Test Automation using UiPath Test Suite series, part 1
UiPath Test Automation using UiPath Test Suite series, part 1DianaGray10
 
Search and Society: Reimagining Information Access for Radical Futures
Search and Society: Reimagining Information Access for Radical FuturesSearch and Society: Reimagining Information Access for Radical Futures
Search and Society: Reimagining Information Access for Radical FuturesBhaskar Mitra
 
Salesforce Adoption – Metrics, Methods, and Motivation, Antone Kom
Salesforce Adoption – Metrics, Methods, and Motivation, Antone KomSalesforce Adoption – Metrics, Methods, and Motivation, Antone Kom
Salesforce Adoption – Metrics, Methods, and Motivation, Antone KomCzechDreamin
 
JMeter webinar - integration with InfluxDB and Grafana
JMeter webinar - integration with InfluxDB and GrafanaJMeter webinar - integration with InfluxDB and Grafana
JMeter webinar - integration with InfluxDB and GrafanaRTTS
 
Knowledge engineering: from people to machines and back
Knowledge engineering: from people to machines and backKnowledge engineering: from people to machines and back
Knowledge engineering: from people to machines and backElena Simperl
 
Powerful Start- the Key to Project Success, Barbara Laskowska
Powerful Start- the Key to Project Success, Barbara LaskowskaPowerful Start- the Key to Project Success, Barbara Laskowska
Powerful Start- the Key to Project Success, Barbara LaskowskaCzechDreamin
 
10 Differences between Sales Cloud and CPQ, Blanka Doktorová
10 Differences between Sales Cloud and CPQ, Blanka Doktorová10 Differences between Sales Cloud and CPQ, Blanka Doktorová
10 Differences between Sales Cloud and CPQ, Blanka DoktorováCzechDreamin
 
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 ThousandEyesThousandEyes
 
Empowering NextGen Mobility via Large Action Model Infrastructure (LAMI): pav...
Empowering NextGen Mobility via Large Action Model Infrastructure (LAMI): pav...Empowering NextGen Mobility via Large Action Model Infrastructure (LAMI): pav...
Empowering NextGen Mobility via Large Action Model Infrastructure (LAMI): pav...Thierry Lestable
 
Connector Corner: Automate dynamic content and events by pushing a button
Connector Corner: Automate dynamic content and events by pushing a buttonConnector Corner: Automate dynamic content and events by pushing a button
Connector Corner: Automate dynamic content and events by pushing a buttonDianaGray10
 
Custom Approval Process: A New Perspective, Pavel Hrbacek & Anindya Halder
Custom Approval Process: A New Perspective, Pavel Hrbacek & Anindya HalderCustom Approval Process: A New Perspective, Pavel Hrbacek & Anindya Halder
Custom Approval Process: A New Perspective, Pavel Hrbacek & Anindya HalderCzechDreamin
 
Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...
Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...
Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...CzechDreamin
 
Slack (or Teams) Automation for Bonterra Impact Management (fka Social Soluti...
Slack (or Teams) Automation for Bonterra Impact Management (fka Social Soluti...Slack (or Teams) Automation for Bonterra Impact Management (fka Social Soluti...
Slack (or Teams) Automation for Bonterra Impact Management (fka Social Soluti...Jeffrey Haguewood
 
Kubernetes & AI - Beauty and the Beast !?! @KCD Istanbul 2024
Kubernetes & AI - Beauty and the Beast !?! @KCD Istanbul 2024Kubernetes & AI - Beauty and the Beast !?! @KCD Istanbul 2024
Kubernetes & AI - Beauty and the Beast !?! @KCD Istanbul 2024Tobias Schneck
 
Essentials of Automations: Optimizing FME Workflows with Parameters
Essentials of Automations: Optimizing FME Workflows with ParametersEssentials of Automations: Optimizing FME Workflows with Parameters
Essentials of Automations: Optimizing FME Workflows with ParametersSafe Software
 
Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...
Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...
Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...UiPathCommunity
 
ODC, Data Fabric and Architecture User Group
ODC, Data Fabric and Architecture User GroupODC, Data Fabric and Architecture User Group
ODC, Data Fabric and Architecture User GroupCatarinaPereira64715
 
From Siloed Products to Connected Ecosystem: Building a Sustainable and Scala...
From Siloed Products to Connected Ecosystem: Building a Sustainable and Scala...From Siloed Products to Connected Ecosystem: Building a Sustainable and Scala...
From Siloed Products to Connected Ecosystem: Building a Sustainable and Scala...Product School
 
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...Product School
 

Recently uploaded (20)

IoT Analytics Company Presentation May 2024
IoT Analytics Company Presentation May 2024IoT Analytics Company Presentation May 2024
IoT Analytics Company Presentation May 2024
 
UiPath Test Automation using UiPath Test Suite series, part 1
UiPath Test Automation using UiPath Test Suite series, part 1UiPath Test Automation using UiPath Test Suite series, part 1
UiPath Test Automation using UiPath Test Suite series, part 1
 
Search and Society: Reimagining Information Access for Radical Futures
Search and Society: Reimagining Information Access for Radical FuturesSearch and Society: Reimagining Information Access for Radical Futures
Search and Society: Reimagining Information Access for Radical Futures
 
Salesforce Adoption – Metrics, Methods, and Motivation, Antone Kom
Salesforce Adoption – Metrics, Methods, and Motivation, Antone KomSalesforce Adoption – Metrics, Methods, and Motivation, Antone Kom
Salesforce Adoption – Metrics, Methods, and Motivation, Antone Kom
 
JMeter webinar - integration with InfluxDB and Grafana
JMeter webinar - integration with InfluxDB and GrafanaJMeter webinar - integration with InfluxDB and Grafana
JMeter webinar - integration with InfluxDB and Grafana
 
Knowledge engineering: from people to machines and back
Knowledge engineering: from people to machines and backKnowledge engineering: from people to machines and back
Knowledge engineering: from people to machines and back
 
Powerful Start- the Key to Project Success, Barbara Laskowska
Powerful Start- the Key to Project Success, Barbara LaskowskaPowerful Start- the Key to Project Success, Barbara Laskowska
Powerful Start- the Key to Project Success, Barbara Laskowska
 
10 Differences between Sales Cloud and CPQ, Blanka Doktorová
10 Differences between Sales Cloud and CPQ, Blanka Doktorová10 Differences between Sales Cloud and CPQ, Blanka Doktorová
10 Differences between Sales Cloud and CPQ, Blanka Doktorová
 
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
 
Empowering NextGen Mobility via Large Action Model Infrastructure (LAMI): pav...
Empowering NextGen Mobility via Large Action Model Infrastructure (LAMI): pav...Empowering NextGen Mobility via Large Action Model Infrastructure (LAMI): pav...
Empowering NextGen Mobility via Large Action Model Infrastructure (LAMI): pav...
 
Connector Corner: Automate dynamic content and events by pushing a button
Connector Corner: Automate dynamic content and events by pushing a buttonConnector Corner: Automate dynamic content and events by pushing a button
Connector Corner: Automate dynamic content and events by pushing a button
 
Custom Approval Process: A New Perspective, Pavel Hrbacek & Anindya Halder
Custom Approval Process: A New Perspective, Pavel Hrbacek & Anindya HalderCustom Approval Process: A New Perspective, Pavel Hrbacek & Anindya Halder
Custom Approval Process: A New Perspective, Pavel Hrbacek & Anindya Halder
 
Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...
Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...
Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...
 
Slack (or Teams) Automation for Bonterra Impact Management (fka Social Soluti...
Slack (or Teams) Automation for Bonterra Impact Management (fka Social Soluti...Slack (or Teams) Automation for Bonterra Impact Management (fka Social Soluti...
Slack (or Teams) Automation for Bonterra Impact Management (fka Social Soluti...
 
Kubernetes & AI - Beauty and the Beast !?! @KCD Istanbul 2024
Kubernetes & AI - Beauty and the Beast !?! @KCD Istanbul 2024Kubernetes & AI - Beauty and the Beast !?! @KCD Istanbul 2024
Kubernetes & AI - Beauty and the Beast !?! @KCD Istanbul 2024
 
Essentials of Automations: Optimizing FME Workflows with Parameters
Essentials of Automations: Optimizing FME Workflows with ParametersEssentials of Automations: Optimizing FME Workflows with Parameters
Essentials of Automations: Optimizing FME Workflows with Parameters
 
Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...
Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...
Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...
 
ODC, Data Fabric and Architecture User Group
ODC, Data Fabric and Architecture User GroupODC, Data Fabric and Architecture User Group
ODC, Data Fabric and Architecture User Group
 
From Siloed Products to Connected Ecosystem: Building a Sustainable and Scala...
From Siloed Products to Connected Ecosystem: Building a Sustainable and Scala...From Siloed Products to Connected Ecosystem: Building a Sustainable and Scala...
From Siloed Products to Connected Ecosystem: Building a Sustainable and Scala...
 
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
 

Computer Vision with Deep Learning

  • 1. Computer Vision with Deep Learning OOP Munich, 07.02.2018, Andreas Eßbaumer
  • 2. Computer Vision with Deep Learning | Andreas Eßbaumer ▪ History and Hype of Deep Learning and Convolution Neural Networks ▪ Visualization of Convolutions ▪ The three and a half types of Computer Vision applications ▪ How to implement Deep Learning for Computer Vision? ▪ Image Classification: Pipelines ▪ Object Detection: Filler caps © 2017 Capgemini. All rights reserved. 2 Overview
  • 3. Computer Vision with Deep Learning | Andreas Eßbaumer 3© 2018 Capgemini. All rights reserved. A bit of history Source: Bennamoun, DeepLearn 2017
  • 4. Computer Vision with Deep Learning | Andreas Eßbaumer 4© 2018 Capgemini. All rights reserved. A bit of history Source: Bennamoun, DeepLearn 2017
  • 5. Computer Vision with Deep Learning | Andreas Eßbaumer Why the hype? – The ImageNet competition ImageNet: Over 10 Mio labeled images. Over 10.000 different labels. ImageNet Competition: Restrict the number of labels to 1000. Algorithms have to produce up to five labels per image.
  • 6. Computer Vision with Deep Learning | Andreas Eßbaumer 6© 2017 Capgemini. All rights reserved. ILSVRC challenge
  • 7. Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 7 Using convolutions the network is able to learn valuable filters and extract the most important features ▪ Each convolutional layer generates many different representations of the image through filters ▪ Filters are in most cases of the sizes 3x3, 5x5 or 7x7 and are learned(!) ▪ Max-pooling areas reduces the size of the image and makes sure filters cover a larger area of the picture ▪ Generally the latter the layer the larger the area of an image which is covered by a filter and the higher the number of filters 10 Layer 1 + 2 3 4 5 6 7 8 9 13 14 15 16 1211 The neural net architecture of VGG 16 Image Source: https://blog.heuritech.com/2016/02/29/a-brief-report-of- the-heuritech-deep-learning-meetup-5/
  • 8. Computer Vision with Deep Learning | Andreas Eßbaumer Visualization of convolution layers learned(!) by the network From millions of training images, the network learns that the most basic features to look for in an image are basic forms like edges, circles or smooth color transitions. Source: Visualizing and Understanding Convolutional Networks by Zeiler/Fergus arXiv:1311.2901
  • 9. Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 9 Visualization of convolution layers learned(!) by the network • Each convolutional layer consists of a high number of filters • Depending on the image some filters are more and differently applicable than others Source: https://github.com/yosinski/deep-visualization-toolbox https://www.youtube.com/watch?v=AgkfIQ4IGaM
  • 10. Computer Vision with Deep Learning | Andreas Eßbaumer Visualization of convolution layers learned(!) by the network Working with the most basic features extracted by the first layers, the network is able to combine them in a larger area of the picture and find more complex structures like certain patterns and written text pieces. Source: Visualizing and Understanding Convolutional Networks by Zeiler/Fergus arXiv:1311.2901
  • 11. Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 11 Visualization of convolution layers learned(!) by the network the number of filters increases to hundreds of specialized filters Source: Visualizing and Understanding Convolutional Networks by Zeiler/Fergus arXiv:1311.2901
  • 12. Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 12 Visualization of convolution layers learned(!) by the network • The filters are purely learned via training data. • There is even a filter highly specialized on detection human faces. Source: https://github.com/yosinski/deep-visualization-toolbox https://www.youtube.com/watch?v=AgkfIQ4IGaM
  • 13. Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 13 There are three and a half types of computer vision applications Image Classification/Recognition Object Detection Image Transformation/FusionDetect similar images
  • 14. Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 14 The two major application types are image classification and object detection Image Classification/Recognition Object Detection • Evaluates the image as a whole • Can categorize different states or even abstract concepts • Looks for physical or at least visually existing objects • Returns multiple objects of various types and with their position
  • 15. Computer Vision with Deep Learning | Andreas Eßbaumer < 15© 2017 Capgemini. All rights reserved.Source: fast.ai • state-of-the-art image classification model in 7 lines of code. • prerequisites are sufficient, representative, labelled training images and GPUs • high-level deep learning framework like Keras make CNNs easy to run • DL-Frameworks like run CNNs efficiently on GPUs. Implementing an Image Classification model is easy (at least easier than it seems)
  • 16. Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 16 To build an object detection model one can use the Tensorflow Object Detection API • No changes on the deep learning code is necessary • the training runs efficiently on GPUs because of Tensorflow • labelling by giving meta information about which types of objects can be found and where (in which box). • Labelling tools are available, e.g. labelImg • Leveraging pre-trained models Useful links: https://github.com/tensorflow/models/tree/master/research/object_detection https://github.com/tzutalin/labelImg
  • 17. Computer Vision with Deep Learning | Andreas Eßbaumer 17© 2017 Capgemini. All rights reserved. • 3.5 million km of pipelines in the world (2014) • Many of these are more than 50 years old. • Pipelines operate under immense pressure: Even small cracks can have disastrous consequences An example: Defect classification for pipeline maintenance • Our client offers solutions for pipeline maintenance • According to regulation pipeline inspections need to be done every few years • pipeline inspection robots for non- destructive testing of pipeline integrity are used
  • 18. Computer Vision with Deep Learning | Andreas Eßbaumer 18© 2017 Capgemini. All rights reserved. • Defect detection is very challenging • Firstly, the sensors produce huge amounts of data that need to be preprocessed and stored online, i.e. during the inspection run. • Secondly, analyzing the images requires special training and is very time intensive. Defect classification with ultrasound • Given an image of a region of interest, can you accurately predict the defect type? The use case A pipeline inspection robot in its rare clean state
  • 19. Computer Vision with Deep Learning | Andreas Eßbaumer Deep Learning gave better results with little effort 0,00% 10,00% 20,00% 30,00% 40,00% 50,00% 60,00% 70,00% 80,00% 90,00% 100,00% CV InceptionV3 - 172+ InceptionV3 - 0+ Custom-V1 Defect classification results Recall Specificity • After the successful POC, industrialization is currently ongoing • Human experts have a strong assistance tool now • Lowering risks by double-checks (Human + AI)
  • 20. Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 20 We use object detection to identify different types of defects and their position on a filler cap • Using a deep learning pipeline of two models to reach full flexibility • The first object detection model runs continuously and looks for various objects e.g. filler caps. • a second object detection model looks for various defects on the surface of the filler cap Model 1 Model 2
  • 21. Computer Vision with Deep Learning | Andreas Eßbaumer © 2017 Capgemini. All rights reserved. 21 Depending on the number and types of defects the quality check is passed or failed • invariant to different environmental lighting, backgrounds, angles, reflections etc. • Detects small defects like scratches, too much paint or punctual damages. • Even the positions of the defects are given by the model providing. ideal information for potential rework • Built in one day • Uses 100 labelled images per filler cap
  • 22. Computer Vision with Deep Learning | Andreas Eßbaumer Andreas Eßbaumer Senior Data Scientist Capgemini Munich Olof-Palme-Straße 14 81829 Munich andreas.essbaumer@capgemini.com M: +49 151 4025 2219 Summary Convolutional Neural Networks (CNNs) • are the State-of-the-Art for Computer Vision • run via Open-Source-Frameworks • require GPUs and labelled Images • are (relatively) easy to run
  • 23. A global leader in consulting, technology services and digital transformation, Capgemini is at the forefront of innovation to address the entire breadth of clients’ opportunities in the evolving world of cloud, digital and platforms. Building on its strong 50-year heritage and deep industry-specific expertise, Capgemini enables organizations to realize their business ambitions through an array of services from strategy to operations. Capgemini is driven by the conviction that the business value of technology comes from and through people. It is a multicultural company of 200,000 team members in over 40 countries. The Group reported 2016 global revenues of EUR 12.5 billion. About Capgemini Learn more about us at www.capgemini.com This message contains information that may be privileged or confidential and is the property of the Capgemini Group. Copyright © 2017 Capgemini. All rights reserved. People matter, results count.