SlideShare une entreprise Scribd logo
1  sur  43
Machine Learning for Mobile Developers:
Tensorflow Lite Framework
Avid Farhoodfar, PhD, MSSW
Artificial intelligence applications in
Consumer Electronic Devices
1
Index
● Why Machine Learning directly on-device is important &
how it is different than what you may do on the server.
● What has been built with TensorFlow Lite.
● Some Demo
2
What is Machine Learning ?
3
What’s a machine learning?
Use algorithms to learn from data (a.k.a Training)
Algorithms are known as models
Models perform prediction (a.k.a inference) Model
Output
“cat”
4
What’s a machine learning?
Use labeled data to improve models
labeled data = Input data + predictions
Errors used to improve the model
We need a Framework to make machine
learning predictions easier
Model
Output
“cat”
“cat”
Error
5
TensorFlow
TensorFlow is google’s framework for
machine learning.
It makes it easy to build and train
neural networks.
It is cross platform, works with CPUs, GPUs,
TPUs, as well as Mobile devices, and
Embedded Platforms.
Model
Output
“cat”
“cat”
Error
6
ML runs in many places
7
ML runs in many places
● Access to more data
8
ML runs in many places
● Access to more data
● Fast and closely knit interactions
9
ML runs in many places
● Access to more data
● Fast and closely knit interactions
● Privacy preserving
10
On-device ML allows building
new types of products!
11
On-device ML is hard
● Produced computer power
12
On-device ML is hard
● Produced computer power
● Limited memory
13
On-device ML is hard
● Produced computer power
● Limited memory
● Battery Constraints
14
Simplifying ML on-device
TensorFlow Lite makes these challenges much easier!
15
What can I do with it?
16
Many use cases
Speech Content
Classification
Prediction
Recognition
Text to Speech
Speech to Text
Object detection
Object Location
OCR Gesture
recognition
Facial modelling
Segmentation
Clustering
Compression
Super Resolution
Translation
Voice Synthesis
Video generation
Text generation
Audio generation
Text Image Audio
17
Who is using it?
18
>2B mobile devices
Have TensorFlow Lite deployed on them in production
19
Some of the users ...
20
21
22
Why Did They Migrate to TFLite?
their
their
23
Where are we at?
Where are we going?
24
25
26
27
28
29
30
GPU vs CPU Performance
At Google, they use
new GPU backend
which is accelerating
compute intensive
networks that enable
vital use cases for
the users.
31
32
33
34
35
36
37
38
39
DEMO TIME
TensorFlow Lite for
Microcontrollers
Smaller, cheaper & wider range of devices
40
This is about
Tiny models on tiny computers!
● Microcontrollers are everywhere
● Speech researchers were pioneers
● Models just tens of kilobytes
41
Here’s one I have in my pocket
Get ready for a live demo!
https://www.sparkfun.com/products/15170
384KB RAM, 1MB Flash, $15
Low single-digit milliwatt power usage
Days on a coin battery!
42
Why is this important?
(1) This is running entirely locally on the embedded chip.
We don’t need to have any internet connection
(2) The model itself is not quite 13 KB but it takes 20KB flash
storage on this device
(3) And the footprint of TensorFlow Lite for
Microcontrollers is only another 25 KB
43

Contenu connexe

Similaire à 2019 04-23-tf lite-avid-f

"Large-Scale Deep Learning for Building Intelligent Computer Systems," a Keyn...
"Large-Scale Deep Learning for Building Intelligent Computer Systems," a Keyn..."Large-Scale Deep Learning for Building Intelligent Computer Systems," a Keyn...
"Large-Scale Deep Learning for Building Intelligent Computer Systems," a Keyn...
Edge AI and Vision Alliance
 
Deep Learning Neural Network Acceleration at the Edge - Andrea Gallo
Deep Learning Neural Network Acceleration at the Edge - Andrea GalloDeep Learning Neural Network Acceleration at the Edge - Andrea Gallo
Deep Learning Neural Network Acceleration at the Edge - Andrea Gallo
Linaro
 
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration WorkflowsIgniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
Safe Software
 
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration Workflows Igniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
Safe Software
 

Similaire à 2019 04-23-tf lite-avid-f (20)

"Large-Scale Deep Learning for Building Intelligent Computer Systems," a Keyn...
"Large-Scale Deep Learning for Building Intelligent Computer Systems," a Keyn..."Large-Scale Deep Learning for Building Intelligent Computer Systems," a Keyn...
"Large-Scale Deep Learning for Building Intelligent Computer Systems," a Keyn...
 
Persian MNIST in 5 Minutes
Persian MNIST in 5 MinutesPersian MNIST in 5 Minutes
Persian MNIST in 5 Minutes
 
Tensorflow Ecosystem
Tensorflow EcosystemTensorflow Ecosystem
Tensorflow Ecosystem
 
Deep learning with TensorFlow
Deep learning with TensorFlowDeep learning with TensorFlow
Deep learning with TensorFlow
 
Machine learning and Deep learning on edge devices using TensorFlow
Machine learning and Deep learning on edge devices using TensorFlowMachine learning and Deep learning on edge devices using TensorFlow
Machine learning and Deep learning on edge devices using TensorFlow
 
Hadoop training in mumbai
Hadoop training in mumbaiHadoop training in mumbai
Hadoop training in mumbai
 
Deep Learning Neural Network Acceleration at the Edge - Andrea Gallo
Deep Learning Neural Network Acceleration at the Edge - Andrea GalloDeep Learning Neural Network Acceleration at the Edge - Andrea Gallo
Deep Learning Neural Network Acceleration at the Edge - Andrea Gallo
 
Kaz Sato, Evangelist, Google at MLconf ATL 2016
Kaz Sato, Evangelist, Google at MLconf ATL 2016Kaz Sato, Evangelist, Google at MLconf ATL 2016
Kaz Sato, Evangelist, Google at MLconf ATL 2016
 
Top 7 Frameworks for Integration AI in App Development
Top 7 Frameworks for Integration AI in App DevelopmentTop 7 Frameworks for Integration AI in App Development
Top 7 Frameworks for Integration AI in App Development
 
2018 09 26 CTT .NET User Group - Introduction to Machine Learning.Net and Win...
2018 09 26 CTT .NET User Group - Introduction to Machine Learning.Net and Win...2018 09 26 CTT .NET User Group - Introduction to Machine Learning.Net and Win...
2018 09 26 CTT .NET User Group - Introduction to Machine Learning.Net and Win...
 
Introduction to ML.NET
Introduction to ML.NETIntroduction to ML.NET
Introduction to ML.NET
 
Martin Musiol - Style-transfer with TensorFlow - Codemotion Milan 2018
Martin Musiol - Style-transfer with TensorFlow - Codemotion Milan 2018Martin Musiol - Style-transfer with TensorFlow - Codemotion Milan 2018
Martin Musiol - Style-transfer with TensorFlow - Codemotion Milan 2018
 
Machine Learning pada Perangkat Portabel - Nurendrantoro
Machine Learning pada Perangkat Portabel - NurendrantoroMachine Learning pada Perangkat Portabel - Nurendrantoro
Machine Learning pada Perangkat Portabel - Nurendrantoro
 
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration WorkflowsIgniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
 
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration Workflows Igniting Next Level Productivity with AI-Infused Data Integration Workflows
Igniting Next Level Productivity with AI-Infused Data Integration Workflows
 
Getting Started with Machine Learning.Net & Windows Machine Learning
Getting Started with Machine Learning.Net & Windows Machine LearningGetting Started with Machine Learning.Net & Windows Machine Learning
Getting Started with Machine Learning.Net & Windows Machine Learning
 
Track2 02. machine intelligence at google scale google, kaz sato, staff devel...
Track2 02. machine intelligence at google scale google, kaz sato, staff devel...Track2 02. machine intelligence at google scale google, kaz sato, staff devel...
Track2 02. machine intelligence at google scale google, kaz sato, staff devel...
 
Leverage the power of machine learning on windows
Leverage the power of machine learning on windowsLeverage the power of machine learning on windows
Leverage the power of machine learning on windows
 
2018 11 14 Artificial Intelligence and Machine Learning in Azure
2018 11 14 Artificial Intelligence and Machine Learning in Azure2018 11 14 Artificial Intelligence and Machine Learning in Azure
2018 11 14 Artificial Intelligence and Machine Learning in Azure
 
Faster deep learning solutions from training to inference - Michele Tameni - ...
Faster deep learning solutions from training to inference - Michele Tameni - ...Faster deep learning solutions from training to inference - Michele Tameni - ...
Faster deep learning solutions from training to inference - Michele Tameni - ...
 

Dernier

Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
ZurliaSoop
 
Salient Features of India constitution especially power and functions
Salient Features of India constitution especially power and functionsSalient Features of India constitution especially power and functions
Salient Features of India constitution especially power and functions
KarakKing
 
Spellings Wk 3 English CAPS CARES Please Practise
Spellings Wk 3 English CAPS CARES Please PractiseSpellings Wk 3 English CAPS CARES Please Practise
Spellings Wk 3 English CAPS CARES Please Practise
AnaAcapella
 

Dernier (20)

Food safety_Challenges food safety laboratories_.pdf
Food safety_Challenges food safety laboratories_.pdfFood safety_Challenges food safety laboratories_.pdf
Food safety_Challenges food safety laboratories_.pdf
 
SOC 101 Demonstration of Learning Presentation
SOC 101 Demonstration of Learning PresentationSOC 101 Demonstration of Learning Presentation
SOC 101 Demonstration of Learning Presentation
 
80 ĐỀ THI THỬ TUYỂN SINH TIẾNG ANH VÀO 10 SỞ GD – ĐT THÀNH PHỐ HỒ CHÍ MINH NĂ...
80 ĐỀ THI THỬ TUYỂN SINH TIẾNG ANH VÀO 10 SỞ GD – ĐT THÀNH PHỐ HỒ CHÍ MINH NĂ...80 ĐỀ THI THỬ TUYỂN SINH TIẾNG ANH VÀO 10 SỞ GD – ĐT THÀNH PHỐ HỒ CHÍ MINH NĂ...
80 ĐỀ THI THỬ TUYỂN SINH TIẾNG ANH VÀO 10 SỞ GD – ĐT THÀNH PHỐ HỒ CHÍ MINH NĂ...
 
FSB Advising Checklist - Orientation 2024
FSB Advising Checklist - Orientation 2024FSB Advising Checklist - Orientation 2024
FSB Advising Checklist - Orientation 2024
 
Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
 
Accessible Digital Futures project (20/03/2024)
Accessible Digital Futures project (20/03/2024)Accessible Digital Futures project (20/03/2024)
Accessible Digital Futures project (20/03/2024)
 
Google Gemini An AI Revolution in Education.pptx
Google Gemini An AI Revolution in Education.pptxGoogle Gemini An AI Revolution in Education.pptx
Google Gemini An AI Revolution in Education.pptx
 
Single or Multiple melodic lines structure
Single or Multiple melodic lines structureSingle or Multiple melodic lines structure
Single or Multiple melodic lines structure
 
How to Create and Manage Wizard in Odoo 17
How to Create and Manage Wizard in Odoo 17How to Create and Manage Wizard in Odoo 17
How to Create and Manage Wizard in Odoo 17
 
General Principles of Intellectual Property: Concepts of Intellectual Proper...
General Principles of Intellectual Property: Concepts of Intellectual  Proper...General Principles of Intellectual Property: Concepts of Intellectual  Proper...
General Principles of Intellectual Property: Concepts of Intellectual Proper...
 
Unit-V; Pricing (Pharma Marketing Management).pptx
Unit-V; Pricing (Pharma Marketing Management).pptxUnit-V; Pricing (Pharma Marketing Management).pptx
Unit-V; Pricing (Pharma Marketing Management).pptx
 
Unit 3 Emotional Intelligence and Spiritual Intelligence.pdf
Unit 3 Emotional Intelligence and Spiritual Intelligence.pdfUnit 3 Emotional Intelligence and Spiritual Intelligence.pdf
Unit 3 Emotional Intelligence and Spiritual Intelligence.pdf
 
Sociology 101 Demonstration of Learning Exhibit
Sociology 101 Demonstration of Learning ExhibitSociology 101 Demonstration of Learning Exhibit
Sociology 101 Demonstration of Learning Exhibit
 
How to setup Pycharm environment for Odoo 17.pptx
How to setup Pycharm environment for Odoo 17.pptxHow to setup Pycharm environment for Odoo 17.pptx
How to setup Pycharm environment for Odoo 17.pptx
 
Salient Features of India constitution especially power and functions
Salient Features of India constitution especially power and functionsSalient Features of India constitution especially power and functions
Salient Features of India constitution especially power and functions
 
HMCS Vancouver Pre-Deployment Brief - May 2024 (Web Version).pptx
HMCS Vancouver Pre-Deployment Brief - May 2024 (Web Version).pptxHMCS Vancouver Pre-Deployment Brief - May 2024 (Web Version).pptx
HMCS Vancouver Pre-Deployment Brief - May 2024 (Web Version).pptx
 
On National Teacher Day, meet the 2024-25 Kenan Fellows
On National Teacher Day, meet the 2024-25 Kenan FellowsOn National Teacher Day, meet the 2024-25 Kenan Fellows
On National Teacher Day, meet the 2024-25 Kenan Fellows
 
Graduate Outcomes Presentation Slides - English
Graduate Outcomes Presentation Slides - EnglishGraduate Outcomes Presentation Slides - English
Graduate Outcomes Presentation Slides - English
 
TỔNG ÔN TẬP THI VÀO LỚP 10 MÔN TIẾNG ANH NĂM HỌC 2023 - 2024 CÓ ĐÁP ÁN (NGỮ Â...
TỔNG ÔN TẬP THI VÀO LỚP 10 MÔN TIẾNG ANH NĂM HỌC 2023 - 2024 CÓ ĐÁP ÁN (NGỮ Â...TỔNG ÔN TẬP THI VÀO LỚP 10 MÔN TIẾNG ANH NĂM HỌC 2023 - 2024 CÓ ĐÁP ÁN (NGỮ Â...
TỔNG ÔN TẬP THI VÀO LỚP 10 MÔN TIẾNG ANH NĂM HỌC 2023 - 2024 CÓ ĐÁP ÁN (NGỮ Â...
 
Spellings Wk 3 English CAPS CARES Please Practise
Spellings Wk 3 English CAPS CARES Please PractiseSpellings Wk 3 English CAPS CARES Please Practise
Spellings Wk 3 English CAPS CARES Please Practise
 

2019 04-23-tf lite-avid-f

  • 1. Machine Learning for Mobile Developers: Tensorflow Lite Framework Avid Farhoodfar, PhD, MSSW Artificial intelligence applications in Consumer Electronic Devices 1
  • 2. Index ● Why Machine Learning directly on-device is important & how it is different than what you may do on the server. ● What has been built with TensorFlow Lite. ● Some Demo 2
  • 3. What is Machine Learning ? 3
  • 4. What’s a machine learning? Use algorithms to learn from data (a.k.a Training) Algorithms are known as models Models perform prediction (a.k.a inference) Model Output “cat” 4
  • 5. What’s a machine learning? Use labeled data to improve models labeled data = Input data + predictions Errors used to improve the model We need a Framework to make machine learning predictions easier Model Output “cat” “cat” Error 5
  • 6. TensorFlow TensorFlow is google’s framework for machine learning. It makes it easy to build and train neural networks. It is cross platform, works with CPUs, GPUs, TPUs, as well as Mobile devices, and Embedded Platforms. Model Output “cat” “cat” Error 6
  • 7. ML runs in many places 7
  • 8. ML runs in many places ● Access to more data 8
  • 9. ML runs in many places ● Access to more data ● Fast and closely knit interactions 9
  • 10. ML runs in many places ● Access to more data ● Fast and closely knit interactions ● Privacy preserving 10
  • 11. On-device ML allows building new types of products! 11
  • 12. On-device ML is hard ● Produced computer power 12
  • 13. On-device ML is hard ● Produced computer power ● Limited memory 13
  • 14. On-device ML is hard ● Produced computer power ● Limited memory ● Battery Constraints 14
  • 15. Simplifying ML on-device TensorFlow Lite makes these challenges much easier! 15
  • 16. What can I do with it? 16
  • 17. Many use cases Speech Content Classification Prediction Recognition Text to Speech Speech to Text Object detection Object Location OCR Gesture recognition Facial modelling Segmentation Clustering Compression Super Resolution Translation Voice Synthesis Video generation Text generation Audio generation Text Image Audio 17
  • 18. Who is using it? 18
  • 19. >2B mobile devices Have TensorFlow Lite deployed on them in production 19
  • 20. Some of the users ... 20
  • 21. 21
  • 22. 22
  • 23. Why Did They Migrate to TFLite? their their 23
  • 24. Where are we at? Where are we going? 24
  • 25. 25
  • 26. 26
  • 27. 27
  • 28. 28
  • 29. 29
  • 30. 30 GPU vs CPU Performance At Google, they use new GPU backend which is accelerating compute intensive networks that enable vital use cases for the users.
  • 31. 31
  • 32. 32
  • 33. 33
  • 34. 34
  • 35. 35
  • 36. 36
  • 37. 37
  • 38. 38
  • 39. 39
  • 40. DEMO TIME TensorFlow Lite for Microcontrollers Smaller, cheaper & wider range of devices 40
  • 41. This is about Tiny models on tiny computers! ● Microcontrollers are everywhere ● Speech researchers were pioneers ● Models just tens of kilobytes 41
  • 42. Here’s one I have in my pocket Get ready for a live demo! https://www.sparkfun.com/products/15170 384KB RAM, 1MB Flash, $15 Low single-digit milliwatt power usage Days on a coin battery! 42
  • 43. Why is this important? (1) This is running entirely locally on the embedded chip. We don’t need to have any internet connection (2) The model itself is not quite 13 KB but it takes 20KB flash storage on this device (3) And the footprint of TensorFlow Lite for Microcontrollers is only another 25 KB 43