KIVI Innovation Drinks Twente An AI encounter with BrainCreators - - PowerPoint PPT Presentation

▶
kivi innovation drinks twente
SMART_READER_LITE
LIVE PREVIEW

KIVI Innovation Drinks Twente An AI encounter with BrainCreators - - PowerPoint PPT Presentation

Mar 29, 2023 118 likes •812 views

KIVI Innovation Drinks Twente An AI encounter with BrainCreators Overview Short introduction BrainCreators Example use-cases AI maturity model The Application Gap Epilogue: pretty pictures BrainCreators applies 20+

slide-1
SLIDE 1

KIVI Innovation Drinks Twente

An AI encounter with BrainCreators

slide-2
SLIDE 2

Overview

➢ Short introduction BrainCreators ➢ Example use-cases ➢ AI maturity model ➢ The Application Gap ➢ Epilogue: pretty pictures

slide-3
SLIDE 3
slide-4
SLIDE 4

Discover value Deploy solutions Accelerate teams

Compile a strategic roadmap of viable business cases Implement scalable solutions with maximum business impact Inherit skills & best practices with expert coaching

BrainCreators applies 20+ years

  • f experience in artificial intelligence to

business challenges across all verticals

slide-5
SLIDE 5

TRUSTED BY

slide-6
SLIDE 6

Use cases

➢ Smart Radio ➢ Logistics ➢ Fashion and Retail ➢ Steel quality control ➢ Genetics ➢ Telecommunication

slide-7
SLIDE 7

Smart Radio

  • 10 hours original radio broadcast per day
  • Additional podcast creation
  • Inconsistent manual tagging of content
  • Course segmentation of topics

24/7 news radio

  • Automatically curated playlists of news content
  • Taylored to a listener’s preferences
  • On demand

Smart Radio

slide-8
SLIDE 8

Smart Radio

  • Audio feature detection
  • Detection of semantic overlap among existing labels
  • Semi-supervised refinement of existing dataset
  • Topic modeling
  • Segment classification

AI under the hood

slide-9
SLIDE 9

Smart Radio

  • Detect topics in segments
  • Cut audio segments
  • Provide relevant user content

Results (v1 in the making)

slide-10
SLIDE 10

Logistics

  • Manual identification of address data for 15% of total volume
  • 4% delivered to wrong address
  • Geographical location of delivery points imprecise
  • Delivery window too coarse

Before application of AI

slide-11
SLIDE 11

Logistics

  • Fuzzy logic address matching
  • GPS delivery point prediction
  • Time window estimation & optimisation
  • Automated location mapping (inc. po-boxes)
  • Trained on historic data and self learning

AI under the hood

slide-12
SLIDE 12

Logistics

  • Manual correction reduced to <2% of total volume
  • Delivery failures reduced by 50%
  • 2000 man hours saved per month
  • Improved customer service through better time windows

Results

slide-13
SLIDE 13

Fashion & retail

  • Manual classification of products
  • Complex mappings to market place taxonomies
  • Poor quality of properties data
  • Basic recommendations

Before

slide-14
SLIDE 14

Fashion & retail

  • Training set of 20+ Million products
  • Combined Image & Text classifiers
  • Sorting of products using complex features
  • Human-in-the-loop data improvement

Under the hood

slide-15
SLIDE 15

Fashion & retail

  • Automated categorization >95% accurate
  • Auto-enrichment of product data
  • Product family recommendation
  • Cross & upselling automation

Result

slide-16
SLIDE 16
  • A major European steel producer
  • Total of 7.1 million tonnes of steel

products in 2016

  • High quality sheet and strip steel
  • Automotive, packaging, and construction

sectors

General

Steel quality control

slide-17
SLIDE 17
  • Kilometers of steel sheet each day
  • Accurate quality assessment enables

more profitable trading

  • Defects need to be detected to prevent

machine breaks

  • Manual inspection supported by

automatic camera system

Initial Situation

Steel quality control

slide-18
SLIDE 18
  • Infrared cameras inspect moving steel

sheets on conveyor belts

  • Basic image processing detects regions
  • f interest
  • Manual inspection often needed
  • Accuracy can still be improved

Camera system

Steel quality control

slide-19
SLIDE 19
  • Up to 50 different defect types
  • 5 million (!) new images each day
  • Currently only 25 thousand annotated

images available in total

  • Severely imbalanced data sets
  • Manual annotation is costly

Data sets

Steel quality control

slide-20
SLIDE 20
  • Deep Learning for robust image

classification

  • Ai & Active Learning approach for

efficient image annotation

  • Integration in existing systems
  • Knowledge transfer to customer’s own

tech team

Solution

Steel quality control

slide-21
SLIDE 21

Genetics for livestock

the animal protein value chain

slide-22
SLIDE 22

Genetics for livestock

the animal protein value chain

slide-23
SLIDE 23
  • Large scale selective breeding as an

industrial optimization process

  • Changing targets due to commercial,

political, and environmental requirements

  • Integration of different data sets,

including genomics data

  • Evaluation is either slow or imprecise

Selective breeding

Genetics for livestock

Breeding value = Genetics + Environment

slide-24
SLIDE 24
  • Predict carcass properties from

measurements on live animals

  • Numerical input data, e.g. weights at

different ages

  • Ultrascan visual data
  • How can the ultrascans be used more

effectively ?

Challenge

Genetics for livestock

slide-25
SLIDE 25
  • Information can get lost in the narrow

passage of human interpretation

  • Deep learning helps to extract useful

information from complex visual data

  • Less requirements for human

understanding of the images

  • End-to-end learning combines visual

and non-visual data into one system

Narrow passage

Genetics for livestock

Human understanding Deep Learning

slide-26
SLIDE 26
  • Very large telco network
  • Hybrid Fibre Coax
  • Up to 5M modems
  • Modems report their status
  • Thousands of relay points
  • Diversity of legacy systems

Initial situation

Fault detection in telecom

…..

slide-27
SLIDE 27
  • Manage fleet of field technicians
  • Network errors and maintenance
  • Detect & classify problems
  • Find problem root causes
  • Collect useful data

Challenge

Fault detection in telecom

slide-28
SLIDE 28
  • Dedicated data labeling software
  • Network Anomaly Detection
  • Generalize to all fault types
  • AI Roadmap

Solution

Fault detection in telecom

Human understanding Deep Learning

slide-29
SLIDE 29

TRANSFORMING TO A DIGITAL ENTERPRISE

slide-30
SLIDE 30

Exploring

TRANSFORMING TO A DIGITAL ENTERPRISE

slide-31
SLIDE 31

Exploring Planning

TRANSFORMING TO A DIGITAL ENTERPRISE

slide-32
SLIDE 32

Exploring Planning Experimenting

TRANSFORMING TO A DIGITAL ENTERPRISE

slide-33
SLIDE 33

Exploring Planning Experimenting Productizing

TRANSFORMING TO A DIGITAL ENTERPRISE

slide-34
SLIDE 34

Exploring Planning Experimenting Productizing Scaling

TRANSFORMING TO A DIGITAL ENTERPRISE

slide-35
SLIDE 35

Exploring Planning Experimenting Productizing Scaling Data-Centric

TRANSFORMING TO A DIGITAL ENTERPRISE

slide-36
SLIDE 36

The Application Gap

slide-37
SLIDE 37

The Application Gap

… between Research and Industry

slide-38
SLIDE 38

The Application Gap

… between Research and Industry

  • When is something “solved” ?
  • Has it been demonstrated to work once,

under special circumstances?

  • Or is it ready and safe to deploy in

general, right now, for everyone ?

  • Speech-to-text ?
  • Self-driving cars ?
  • … …

Solved? .... really?

slide-39
SLIDE 39

Andrew Ng: “AI is the new electricity”

“We have enough papers. Stop publishing, and start transforming people’s lives with technology!”

slide-40
SLIDE 40

The competitive landscape

slide-41
SLIDE 41

Thank you!

Maarten Stol

maarten.stol@braincreators.com

BrainCreators

Prinsengracht 697 1017JV Amsterdam +31 (0)20 369 7260

slide-42
SLIDE 42

Epilogue

slide-43
SLIDE 43

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-44
SLIDE 44

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-45
SLIDE 45

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-46
SLIDE 46

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-47
SLIDE 47

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-48
SLIDE 48

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-49
SLIDE 49

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-50
SLIDE 50

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-51
SLIDE 51

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-52
SLIDE 52

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-53
SLIDE 53

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-54
SLIDE 54

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-55
SLIDE 55

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-56
SLIDE 56

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-57
SLIDE 57

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-58
SLIDE 58

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-59
SLIDE 59

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-60
SLIDE 60

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-61
SLIDE 61

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-62
SLIDE 62

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-63
SLIDE 63

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-64
SLIDE 64

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-65
SLIDE 65

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-66
SLIDE 66

BigGAN

Large Scale GAN Training for High Fidelity Natural Image Synthesis Andrew Brock, Jeff Donahue, Karen Simonyan

slide-67
SLIDE 67

GAN Progress

Source: www.eff.org/files/2018/02/20/malicious_ai_report_final.pdf

slide-68
SLIDE 68

Source: PROGRESSIVE GROWING OF GANS FOR IMPROVED QUALITY, STABILITY, AND VARIATION Tero Karras, et.al. NVIDIA

slide-69
SLIDE 69

Source: PROGRESSIVE GROWING OF GANS FOR IMPROVED QUALITY, STABILITY, AND VARIATION Tero Karras, et.al. NVIDIA