SlideShare ist ein Scribd-Unternehmen logo
1 von 27
1
Confidential
AIoT from Cloud to Fog Computing
Dov Nimratz
Senior Solution Architect
November 2019
2
Confidential
2
Agenda
1. About Myself
2. Historical Content
3. Why now
4. IoT and AI synergy
5. AIoT Edge and Cloud Accelerator Parts
6. What to develop
Confidential
3
About Myself
Dov Nimartz
● 30+ years in R&D
● 17 years in Israel HighTech ECI, Telrad, RAD, Audiocodes
companies
● HW, SW, Mechanical design engineer
● Project & Product Manager
● Business developer for EMEA & CIS
● Solution Architect
● 30+ publications, US patent, 70+ speeches
● Counseling & SW development teaching
dov.nimratz@globallogic.com
Skype: dovnmr
Confidential
4
Historical Content
Analog
processors
Domain specific
models
CPU
computing
GPU
computing
VPU
computing
TPU
computing
CNNNN Optimizations
Confidential
5
Why Now
Intel Neural Network Stick
• Price $109
• Size 75x27x14 mm
• Weight 18 gram
• Based on Myriad X VPU
• 16 video CPU
• MobileNet v2 20 fps
• TensorFlow, Cafee
• Current up to 40 mA
NVIDIA Jetson Nano
• Price $100
• Size 95x76x30mm
• Weight 136 gram
• 64bit quad-core A57
• 128-core Nvidia GPU
• MobileNet v2 12 fps
• Current up to 1220 mA
Google TPU Accelerator
• Price $75
• Size 65x30x8 mm
• Weight 12 gram
• Google Edge TPU
• MobileNet v2 40+ fps
• TensorFlow Lite
• Current up to 20 mA
Google Coral Dev Board
• Price $150
• Size 85x56x20mm
• Weight 78 gram
• NXP i.MX 8M SOC
• Google Edge TPU
• MobileNet v2 100+ fps
• TensorFlow Lite
• Current up to 960 mA
Confidential
6
Platforms Benchmark
Resume:
• Absolute winner is Coral Dev Board.
• Intel NN Stick is effective at face
recognition related projects.
• Coral USB Accelerator is the cheapest
solution with a good performance.
• NVIDIA Jetson Nano is not
recommended for use.
Confidential
7
IoT and AI Synergy
Collect
StoreExecute
Analyse Process
IoT
Collect
StoreExecute
Analyse Process
AI
8
Confidential
8
Edge IoT System Concept
Cloud part consists of:
● different but similar components
● data ETL core
● different presentation layers.
But, we may have similar sensors or
Edge Layer:
AI Embedded in the Edge
Physical
Edge
Local
Network
Gateway
Wide
Network
Security
Middleware
ETL
Presentation
Notification Configuration
BackEndEdge
Ad Hoc /
Mesh
M2MM2P
Big Data,
Analytic
9
Confidential
9
Edge Layer
Watchdog
State
Mashine
Main
Logic
Sensor
Configuration
Collected
data
Local
Network
Layer
Battery
Edge Layer Edge Part
First business layer responsible for:
● Gather information
● Abstraction generation
● Neighbors redundancy communication
10
Confidential
10
Middleware Layer
Load
Balancers
Queues
Subscribers
Middleware Part
● Spread information by load
● Provide asynchronous call
computation
11
Confidential
11
ETL Layer
Extract Transform
ArchiveBucket
OLAP
OLTP
SQL
Notifications
If the data does not have a strict scheme
(table columns), then it is stored in NoSQL
databases.
SQL databases have two types:
1. OLTP (Online Transactional Processing)
2. OLAP (Online Analytical Processing)
NoSQL
12
Confidential
12
Edge IoT System Concept
Cloud part consists of
● different but similar components
● data ETL core
● different presentation layers.
But, we may have similar sensors or
Edge Layer:
AI Embedded in the Edge
Physical
Edge
Local
Network
Gateway
Wide
Network
Security
Middleware
ETL
Presentation
Notification Configuration
BackEndEdge
Ad Hoc /
Mesh
M2MM2P
Big Data,
Analytic
13
Confidential
13
Edge AIoT System Concept
Changes from IoT model
● Universal AI Edge sensor
● ML models DB
● IFTTT State machine (Configurable)
● Additional parameters in LvM2M
modes
AI Edge
Local
Network
Gateway
Wide
Network
Security
Middleware
ETL
Presentation
Notification
Configuration
BackEndEdge
Ad Hoc /
Mesh
M2MM2P
FSM
AI Edge
ML Models
IFTTT
14
Confidential
14
Fog Computing AI-Based Architecture Flow
...
...
...
Abstract data
New ML model
Raw data request
and send
Model tuning
● Exchanging knowledge in place of data
● Uploading abstract
● Downloading knowledge
● Self improvement
● Raw data upload by a request from the
upper layer
15
Confidential
15
ML Models Optimization for Edge AI
Frameworks:
● TensorFlow
● TensorFlow Lite
● Caffe
Optimizations:
● Quantization
● Tensor Fusion
MobileNet SSD
SqueezeNet
Inception
Models:
● MobileNet SSD v1 & v2
● SqueezeNet 1.0, SqueezeNet 1.1
● Inception v1 & v2 & v3 3 & v4
Libraries:
● OpenCV
● OpenVINO
QuantizationTensorRT Optimisation
next input
3x3 CBR 3x3 CBR 1x1 CBR
1x1 CBR max pool
next input
input
tensor
Expansion
layer
Depthwise
layer
filter
the
data
Projection
layer
uncompress
the data
compress
the data
outpt
tensor
Min
Input (float)
Min Max
Quantize
QuantizedRelu
Dequantize
Output (float)
Ma
x
Eight
Bit
Min Ma
x
Eight
Bit
maxpool/2
conv1
fire2
fire3
fire4
fire5
fire6
fire7
fire8
fire9
fire10
softmax
maxpool/2
globalavgpool
maxpool/2
96
128
128
256
256
384
384
512
512
1000
16
Confidential
16
IFTTT Final State Machine
● Configurable FSM
● Generic data concept
● Visual simulation
● Visual monitoring
States list
Objects list
Events list
Action list
17
Confidential
AIoT Edge and Cloud Accelerator
Parts
18
Confidential
18
Edge AIoT Accelerator Module
Battery
AI Device
Display
Camera
● A fully standalone video processing
device
● Does not require cloud or mobile device
connection for operation presenting
● Contains replaceable parts
● Single hardware for any IoT solution
demonstration
● Contains port for extended sensors and
Ethernet with PoE
● ML models can be downloaded or
updated from the cloud to support a
specific IoT solution demonstration
19
Confidential
19
AIoT Accelerator Cloud Part
Reaction
Objects
AI Camera
Vision
AB SM
AI camera classifies objects
● Object attributes:
○ ID, probability, location, and state
● AB SM events:
○ Object ID change
○ Time of the day
○ Calendar
○ Public events
● AB SM reactions:
○ Change of state
○ Giving out notifications
○ Control of accelerators
○ Change Object attributes
20
Confidential
20
AIoT Accelerator Block Diagram
Object
Action
Location
Detection Event generation
FSM updateActions
Actions
Internal
● Update Edge ML nodel
● Start record stream
● Take an image shoot
● Assign ID to object/location
● Change object attribute
External
● Make a call
● Control switch/attenuator
● Send a sound message
● ETL action
21
Confidential
What to Develop
22
ConfidentialConfidential
ML Models
Confidential
23
LvM2M Data Model Upgrade
● Objects array
● Probability information
● Correlation data
● Neighbor sensor data
Confidential
24
IFTTT STM Design
● Templates from different solution
● Custom Customization
● Visual simulation
● Visual history and action logic
● Ability for a third party device integration
● API for the external applications
25
Confidential
Questions?
Dov Nimratz
Senior Solution Architect
dov.nimratz@globallogic.com
+380-91-333-3300
26
Confidential
Server
Client
Demonstration
27
Confidential
Thank you
Dov Nimratz
Senior Solution Architect
dov.nimratz@globallogic.com
+380-91-333-3300

Weitere ähnliche Inhalte

Was ist angesagt?

How Cisco Provides World-Class Technology Conference Experiences Using Automa...
How Cisco Provides World-Class Technology Conference Experiences Using Automa...How Cisco Provides World-Class Technology Conference Experiences Using Automa...
How Cisco Provides World-Class Technology Conference Experiences Using Automa...
InfluxData
 
“IoT and Vision: Why It’s a Security Minefield and How to Navigate It,” a Pre...
“IoT and Vision: Why It’s a Security Minefield and How to Navigate It,” a Pre...“IoT and Vision: Why It’s a Security Minefield and How to Navigate It,” a Pre...
“IoT and Vision: Why It’s a Security Minefield and How to Navigate It,” a Pre...
Edge AI and Vision Alliance
 

Was ist angesagt? (20)

Internet of Things - Advantech IoT Gateway Starter Kit
Internet of Things - Advantech IoT Gateway Starter KitInternet of Things - Advantech IoT Gateway Starter Kit
Internet of Things - Advantech IoT Gateway Starter Kit
 
Project Trillium: Arm Machine Learning Platform
Project Trillium: Arm Machine Learning PlatformProject Trillium: Arm Machine Learning Platform
Project Trillium: Arm Machine Learning Platform
 
Developing Robust IoT Gateway Applications from Building Blocks
Developing Robust IoT Gateway Applications from Building BlocksDeveloping Robust IoT Gateway Applications from Building Blocks
Developing Robust IoT Gateway Applications from Building Blocks
 
Arm DynamIQ: Intelligent Solutions Using Cluster Based Multiprocessing
Arm DynamIQ: Intelligent Solutions Using Cluster Based MultiprocessingArm DynamIQ: Intelligent Solutions Using Cluster Based Multiprocessing
Arm DynamIQ: Intelligent Solutions Using Cluster Based Multiprocessing
 
#OSSPARIS19 : RIOT: towards open source, secure DevOps on microcontroller-bas...
#OSSPARIS19 : RIOT: towards open source, secure DevOps on microcontroller-bas...#OSSPARIS19 : RIOT: towards open source, secure DevOps on microcontroller-bas...
#OSSPARIS19 : RIOT: towards open source, secure DevOps on microcontroller-bas...
 
Edge Computing Architecture using GPUs and Kubernetes
Edge Computing Architecture using GPUs and KubernetesEdge Computing Architecture using GPUs and Kubernetes
Edge Computing Architecture using GPUs and Kubernetes
 
How Cisco Provides World-Class Technology Conference Experiences Using Automa...
How Cisco Provides World-Class Technology Conference Experiences Using Automa...How Cisco Provides World-Class Technology Conference Experiences Using Automa...
How Cisco Provides World-Class Technology Conference Experiences Using Automa...
 
IoT Seminar (Oct. 2016) CK Vishwakarma - IOTSG
IoT Seminar (Oct. 2016) CK Vishwakarma - IOTSGIoT Seminar (Oct. 2016) CK Vishwakarma - IOTSG
IoT Seminar (Oct. 2016) CK Vishwakarma - IOTSG
 
Scalable AI Solution cross AI platforms
Scalable AI Solution cross AI platformsScalable AI Solution cross AI platforms
Scalable AI Solution cross AI platforms
 
AI +電腦視覺+嵌入式系統
AI +電腦視覺+嵌入式系統AI +電腦視覺+嵌入式系統
AI +電腦視覺+嵌入式系統
 
Brian Gilmore [InfluxData] | InfluxDB in an IoT Application Architecture | In...
Brian Gilmore [InfluxData] | InfluxDB in an IoT Application Architecture | In...Brian Gilmore [InfluxData] | InfluxDB in an IoT Application Architecture | In...
Brian Gilmore [InfluxData] | InfluxDB in an IoT Application Architecture | In...
 
Airport as a Smartcity - Naftemporiki IoT 2016
Airport as a Smartcity - Naftemporiki IoT 2016Airport as a Smartcity - Naftemporiki IoT 2016
Airport as a Smartcity - Naftemporiki IoT 2016
 
IoTlondon - mbed based IoT Gateway talk
IoTlondon - mbed based IoT Gateway talkIoTlondon - mbed based IoT Gateway talk
IoTlondon - mbed based IoT Gateway talk
 
Thingsboard Gateway - Remote temperature monitoring tool for Industries
Thingsboard Gateway - Remote temperature monitoring tool for IndustriesThingsboard Gateway - Remote temperature monitoring tool for Industries
Thingsboard Gateway - Remote temperature monitoring tool for Industries
 
M-CORD cloud-native
M-CORD cloud-nativeM-CORD cloud-native
M-CORD cloud-native
 
JVM-Con 2017 – Java and IoT, will it blend?
JVM-Con 2017 – Java and IoT, will it blend?JVM-Con 2017 – Java and IoT, will it blend?
JVM-Con 2017 – Java and IoT, will it blend?
 
“IoT and Vision: Why It’s a Security Minefield and How to Navigate It,” a Pre...
“IoT and Vision: Why It’s a Security Minefield and How to Navigate It,” a Pre...“IoT and Vision: Why It’s a Security Minefield and How to Navigate It,” a Pre...
“IoT and Vision: Why It’s a Security Minefield and How to Navigate It,” a Pre...
 
How to Manage Your Time Series Data Pipeline at the Edge with InfluxDB
How to Manage Your Time Series Data Pipeline at the Edge with InfluxDBHow to Manage Your Time Series Data Pipeline at the Edge with InfluxDB
How to Manage Your Time Series Data Pipeline at the Edge with InfluxDB
 
Hyperscalable Unified IoT Platform - Janko Isidorovic - Codemotion Amsterdam ...
Hyperscalable Unified IoT Platform - Janko Isidorovic - Codemotion Amsterdam ...Hyperscalable Unified IoT Platform - Janko Isidorovic - Codemotion Amsterdam ...
Hyperscalable Unified IoT Platform - Janko Isidorovic - Codemotion Amsterdam ...
 
Building an IoT Monitoring App with InfluxDB and LoRa
Building an IoT Monitoring App with InfluxDB and LoRaBuilding an IoT Monitoring App with InfluxDB and LoRa
Building an IoT Monitoring App with InfluxDB and LoRa
 

Ähnlich wie Embedded Fest 2019. Dov Nimratz. Artificial Intelligence in Small Embedded Systems

Tackling Retail Technology Management Challenges at the Edge
Tackling Retail Technology Management Challenges at the EdgeTackling Retail Technology Management Challenges at the Edge
Tackling Retail Technology Management Challenges at the Edge
Rebekah Rodriguez
 
“Making Edge AI Inference Programming Easier and Flexible,” a Presentation fr...
“Making Edge AI Inference Programming Easier and Flexible,” a Presentation fr...“Making Edge AI Inference Programming Easier and Flexible,” a Presentation fr...
“Making Edge AI Inference Programming Easier and Flexible,” a Presentation fr...
Edge AI and Vision Alliance
 

Ähnlich wie Embedded Fest 2019. Dov Nimratz. Artificial Intelligence in Small Embedded Systems (20)

Intels presentation at blue line industrial computer seminar
Intels presentation at blue line industrial computer seminarIntels presentation at blue line industrial computer seminar
Intels presentation at blue line industrial computer seminar
 
NGIoT standardisation workshops_Jens Hagemeyer presentation
NGIoT standardisation workshops_Jens Hagemeyer presentationNGIoT standardisation workshops_Jens Hagemeyer presentation
NGIoT standardisation workshops_Jens Hagemeyer presentation
 
Intel IoT Edge Computing 在 AI 領域的應用與商機
Intel IoT Edge Computing 在 AI 領域的應用與商機Intel IoT Edge Computing 在 AI 領域的應用與商機
Intel IoT Edge Computing 在 AI 領域的應用與商機
 
Harnessing the virtual realm for successful real world artificial intelligence
Harnessing the virtual realm for successful real world artificial intelligenceHarnessing the virtual realm for successful real world artificial intelligence
Harnessing the virtual realm for successful real world artificial intelligence
 
LAS16 100 K1 - Keynote George Grey
LAS16 100 K1 - Keynote George GreyLAS16 100 K1 - Keynote George Grey
LAS16 100 K1 - Keynote George Grey
 
LAS16-100K1: Welcome Keynote
LAS16-100K1: Welcome KeynoteLAS16-100K1: Welcome Keynote
LAS16-100K1: Welcome Keynote
 
Accelerating the AIoT @ the EDGE
Accelerating the AIoT @ the EDGE Accelerating the AIoT @ the EDGE
Accelerating the AIoT @ the EDGE
 
Accelerating Innovation from Edge to Cloud
Accelerating Innovation from Edge to CloudAccelerating Innovation from Edge to Cloud
Accelerating Innovation from Edge to Cloud
 
Integrate the AWS Cloud with Responsive Xilinx Machine Learning at the Edge (...
Integrate the AWS Cloud with Responsive Xilinx Machine Learning at the Edge (...Integrate the AWS Cloud with Responsive Xilinx Machine Learning at the Edge (...
Integrate the AWS Cloud with Responsive Xilinx Machine Learning at the Edge (...
 
Tackling Retail Technology Management Challenges at the Edge
Tackling Retail Technology Management Challenges at the EdgeTackling Retail Technology Management Challenges at the Edge
Tackling Retail Technology Management Challenges at the Edge
 
White Box Hardware Challenges in the 5G & IoT Hyperconnected Era
White Box Hardware Challenges in the 5G & IoT Hyperconnected EraWhite Box Hardware Challenges in the 5G & IoT Hyperconnected Era
White Box Hardware Challenges in the 5G & IoT Hyperconnected Era
 
AIoT: Intelligence on Microcontroller
AIoT: Intelligence on MicrocontrollerAIoT: Intelligence on Microcontroller
AIoT: Intelligence on Microcontroller
 
Dov Nimratz, Roman Chobik "Embedded artificial intelligence"
Dov Nimratz, Roman Chobik "Embedded artificial intelligence"Dov Nimratz, Roman Chobik "Embedded artificial intelligence"
Dov Nimratz, Roman Chobik "Embedded artificial intelligence"
 
HKG18-100K1 - George Grey: Opening Keynote
HKG18-100K1 - George Grey: Opening KeynoteHKG18-100K1 - George Grey: Opening Keynote
HKG18-100K1 - George Grey: Opening Keynote
 
IT Solution through IoT Development
IT Solution through IoT DevelopmentIT Solution through IoT Development
IT Solution through IoT Development
 
General io t_concepts
General io t_conceptsGeneral io t_concepts
General io t_concepts
 
Creating a successful IoT product with MediaTek Labs
Creating a successful IoT product with MediaTek LabsCreating a successful IoT product with MediaTek Labs
Creating a successful IoT product with MediaTek Labs
 
BKK16-500K2 CTO talk - The End to End Story
BKK16-500K2 CTO talk - The End to End StoryBKK16-500K2 CTO talk - The End to End Story
BKK16-500K2 CTO talk - The End to End Story
 
“Making Edge AI Inference Programming Easier and Flexible,” a Presentation fr...
“Making Edge AI Inference Programming Easier and Flexible,” a Presentation fr...“Making Edge AI Inference Programming Easier and Flexible,” a Presentation fr...
“Making Edge AI Inference Programming Easier and Flexible,” a Presentation fr...
 
SS-CPSIoT 2023_Kevin Mika and Piotr Zierhoffer presentation
SS-CPSIoT 2023_Kevin Mika and Piotr Zierhoffer presentationSS-CPSIoT 2023_Kevin Mika and Piotr Zierhoffer presentation
SS-CPSIoT 2023_Kevin Mika and Piotr Zierhoffer presentation
 

Mehr von EmbeddedFest

Mehr von EmbeddedFest (9)

Embedded Fest 2019. Wei Fu. Linux on RISC-V--Fedora and Firmware in practice
Embedded Fest 2019. Wei Fu. Linux on RISC-V--Fedora and Firmware in practiceEmbedded Fest 2019. Wei Fu. Linux on RISC-V--Fedora and Firmware in practice
Embedded Fest 2019. Wei Fu. Linux on RISC-V--Fedora and Firmware in practice
 
Embedded Fest 2019. Руслан Биловол. Linux Boot: The Big Bang theory
Embedded Fest 2019. Руслан Биловол. Linux Boot: The Big Bang theoryEmbedded Fest 2019. Руслан Биловол. Linux Boot: The Big Bang theory
Embedded Fest 2019. Руслан Биловол. Linux Boot: The Big Bang theory
 
Embedded Fest 2019. Віталій Нужний. The Mobility Revolution: the Software tha...
Embedded Fest 2019. Віталій Нужний. The Mobility Revolution: the Software tha...Embedded Fest 2019. Віталій Нужний. The Mobility Revolution: the Software tha...
Embedded Fest 2019. Віталій Нужний. The Mobility Revolution: the Software tha...
 
Embedded Fest 2019. Игорь Опанюк. Das U-boot v2019: a look under the hood
Embedded Fest 2019. Игорь Опанюк. Das U-boot v2019: a look under the hoodEmbedded Fest 2019. Игорь Опанюк. Das U-boot v2019: a look under the hood
Embedded Fest 2019. Игорь Опанюк. Das U-boot v2019: a look under the hood
 
Embedded Fest 2019. Константин Белоусов. Исключения и прерывания на amd64: ка...
Embedded Fest 2019. Константин Белоусов. Исключения и прерывания на amd64: ка...Embedded Fest 2019. Константин Белоусов. Исключения и прерывания на amd64: ка...
Embedded Fest 2019. Константин Белоусов. Исключения и прерывания на amd64: ка...
 
Embedded Fest 2019. Володимир Шанойло. High FIVE: Samsung integrity protectio...
Embedded Fest 2019. Володимир Шанойло. High FIVE: Samsung integrity protectio...Embedded Fest 2019. Володимир Шанойло. High FIVE: Samsung integrity protectio...
Embedded Fest 2019. Володимир Шанойло. High FIVE: Samsung integrity protectio...
 
Embedded Fest 2019. Антон Волошин. Connected Mobility: from Vehicle to Cloud
Embedded Fest 2019. Антон Волошин. Connected Mobility: from Vehicle to CloudEmbedded Fest 2019. Антон Волошин. Connected Mobility: from Vehicle to Cloud
Embedded Fest 2019. Антон Волошин. Connected Mobility: from Vehicle to Cloud
 
Embedded Fest 2019. Игорь Таненков и Игорь Успеньев. Action Recognition from ...
Embedded Fest 2019. Игорь Таненков и Игорь Успеньев. Action Recognition from ...Embedded Fest 2019. Игорь Таненков и Игорь Успеньев. Action Recognition from ...
Embedded Fest 2019. Игорь Таненков и Игорь Успеньев. Action Recognition from ...
 
Embedded Fest 2019. Іван Пустовіт. From AOSP to Android powered device
Embedded Fest 2019. Іван Пустовіт. From AOSP to Android powered deviceEmbedded Fest 2019. Іван Пустовіт. From AOSP to Android powered device
Embedded Fest 2019. Іван Пустовіт. From AOSP to Android powered device
 

Kürzlich hochgeladen

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
 

Kürzlich hochgeladen (20)

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
 
ICT Role in 21st Century Education & its Challenges.pptx
ICT Role in 21st Century Education & its Challenges.pptxICT Role in 21st Century Education & its Challenges.pptx
ICT Role in 21st Century Education & its Challenges.pptx
 
Beyond_Borders_Understanding_Anime_and_Manga_Fandom_A_Comprehensive_Audience_...
Beyond_Borders_Understanding_Anime_and_Manga_Fandom_A_Comprehensive_Audience_...Beyond_Borders_Understanding_Anime_and_Manga_Fandom_A_Comprehensive_Audience_...
Beyond_Borders_Understanding_Anime_and_Manga_Fandom_A_Comprehensive_Audience_...
 
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
 
Graduate Outcomes Presentation Slides - English
Graduate Outcomes Presentation Slides - EnglishGraduate Outcomes Presentation Slides - English
Graduate Outcomes Presentation Slides - English
 
How to Give a Domain for a Field in Odoo 17
How to Give a Domain for a Field in Odoo 17How to Give a Domain for a Field in Odoo 17
How to Give a Domain for a Field in Odoo 17
 
This PowerPoint helps students to consider the concept of infinity.
This PowerPoint helps students to consider the concept of infinity.This PowerPoint helps students to consider the concept of infinity.
This PowerPoint helps students to consider the concept of infinity.
 
Mehran University Newsletter Vol-X, Issue-I, 2024
Mehran University Newsletter Vol-X, Issue-I, 2024Mehran University Newsletter Vol-X, Issue-I, 2024
Mehran University Newsletter Vol-X, Issue-I, 2024
 
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
 
How to Manage Global Discount in Odoo 17 POS
How to Manage Global Discount in Odoo 17 POSHow to Manage Global Discount in Odoo 17 POS
How to Manage Global Discount in Odoo 17 POS
 
Kodo Millet PPT made by Ghanshyam bairwa college of Agriculture kumher bhara...
Kodo Millet  PPT made by Ghanshyam bairwa college of Agriculture kumher bhara...Kodo Millet  PPT made by Ghanshyam bairwa college of Agriculture kumher bhara...
Kodo Millet PPT made by Ghanshyam bairwa college of Agriculture kumher bhara...
 
Sociology 101 Demonstration of Learning Exhibit
Sociology 101 Demonstration of Learning ExhibitSociology 101 Demonstration of Learning Exhibit
Sociology 101 Demonstration of Learning Exhibit
 
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
 
Jamworks pilot and AI at Jisc (20/03/2024)
Jamworks pilot and AI at Jisc (20/03/2024)Jamworks pilot and AI at Jisc (20/03/2024)
Jamworks pilot and AI at Jisc (20/03/2024)
 
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
 
2024-NATIONAL-LEARNING-CAMP-AND-OTHER.pptx
2024-NATIONAL-LEARNING-CAMP-AND-OTHER.pptx2024-NATIONAL-LEARNING-CAMP-AND-OTHER.pptx
2024-NATIONAL-LEARNING-CAMP-AND-OTHER.pptx
 
ICT role in 21st century education and it's challenges.
ICT role in 21st century education and it's challenges.ICT role in 21st century education and it's challenges.
ICT role in 21st century education and it's challenges.
 
Fostering Friendships - Enhancing Social Bonds in the Classroom
Fostering Friendships - Enhancing Social Bonds  in the ClassroomFostering Friendships - Enhancing Social Bonds  in the Classroom
Fostering Friendships - Enhancing Social Bonds in the Classroom
 
Understanding Accommodations and Modifications
Understanding  Accommodations and ModificationsUnderstanding  Accommodations and Modifications
Understanding Accommodations and Modifications
 

Embedded Fest 2019. Dov Nimratz. Artificial Intelligence in Small Embedded Systems

  • 1. 1 Confidential AIoT from Cloud to Fog Computing Dov Nimratz Senior Solution Architect November 2019
  • 2. 2 Confidential 2 Agenda 1. About Myself 2. Historical Content 3. Why now 4. IoT and AI synergy 5. AIoT Edge and Cloud Accelerator Parts 6. What to develop
  • 3. Confidential 3 About Myself Dov Nimartz ● 30+ years in R&D ● 17 years in Israel HighTech ECI, Telrad, RAD, Audiocodes companies ● HW, SW, Mechanical design engineer ● Project & Product Manager ● Business developer for EMEA & CIS ● Solution Architect ● 30+ publications, US patent, 70+ speeches ● Counseling & SW development teaching dov.nimratz@globallogic.com Skype: dovnmr
  • 5. Confidential 5 Why Now Intel Neural Network Stick • Price $109 • Size 75x27x14 mm • Weight 18 gram • Based on Myriad X VPU • 16 video CPU • MobileNet v2 20 fps • TensorFlow, Cafee • Current up to 40 mA NVIDIA Jetson Nano • Price $100 • Size 95x76x30mm • Weight 136 gram • 64bit quad-core A57 • 128-core Nvidia GPU • MobileNet v2 12 fps • Current up to 1220 mA Google TPU Accelerator • Price $75 • Size 65x30x8 mm • Weight 12 gram • Google Edge TPU • MobileNet v2 40+ fps • TensorFlow Lite • Current up to 20 mA Google Coral Dev Board • Price $150 • Size 85x56x20mm • Weight 78 gram • NXP i.MX 8M SOC • Google Edge TPU • MobileNet v2 100+ fps • TensorFlow Lite • Current up to 960 mA
  • 6. Confidential 6 Platforms Benchmark Resume: • Absolute winner is Coral Dev Board. • Intel NN Stick is effective at face recognition related projects. • Coral USB Accelerator is the cheapest solution with a good performance. • NVIDIA Jetson Nano is not recommended for use.
  • 7. Confidential 7 IoT and AI Synergy Collect StoreExecute Analyse Process IoT Collect StoreExecute Analyse Process AI
  • 8. 8 Confidential 8 Edge IoT System Concept Cloud part consists of: ● different but similar components ● data ETL core ● different presentation layers. But, we may have similar sensors or Edge Layer: AI Embedded in the Edge Physical Edge Local Network Gateway Wide Network Security Middleware ETL Presentation Notification Configuration BackEndEdge Ad Hoc / Mesh M2MM2P Big Data, Analytic
  • 9. 9 Confidential 9 Edge Layer Watchdog State Mashine Main Logic Sensor Configuration Collected data Local Network Layer Battery Edge Layer Edge Part First business layer responsible for: ● Gather information ● Abstraction generation ● Neighbors redundancy communication
  • 10. 10 Confidential 10 Middleware Layer Load Balancers Queues Subscribers Middleware Part ● Spread information by load ● Provide asynchronous call computation
  • 11. 11 Confidential 11 ETL Layer Extract Transform ArchiveBucket OLAP OLTP SQL Notifications If the data does not have a strict scheme (table columns), then it is stored in NoSQL databases. SQL databases have two types: 1. OLTP (Online Transactional Processing) 2. OLAP (Online Analytical Processing) NoSQL
  • 12. 12 Confidential 12 Edge IoT System Concept Cloud part consists of ● different but similar components ● data ETL core ● different presentation layers. But, we may have similar sensors or Edge Layer: AI Embedded in the Edge Physical Edge Local Network Gateway Wide Network Security Middleware ETL Presentation Notification Configuration BackEndEdge Ad Hoc / Mesh M2MM2P Big Data, Analytic
  • 13. 13 Confidential 13 Edge AIoT System Concept Changes from IoT model ● Universal AI Edge sensor ● ML models DB ● IFTTT State machine (Configurable) ● Additional parameters in LvM2M modes AI Edge Local Network Gateway Wide Network Security Middleware ETL Presentation Notification Configuration BackEndEdge Ad Hoc / Mesh M2MM2P FSM AI Edge ML Models IFTTT
  • 14. 14 Confidential 14 Fog Computing AI-Based Architecture Flow ... ... ... Abstract data New ML model Raw data request and send Model tuning ● Exchanging knowledge in place of data ● Uploading abstract ● Downloading knowledge ● Self improvement ● Raw data upload by a request from the upper layer
  • 15. 15 Confidential 15 ML Models Optimization for Edge AI Frameworks: ● TensorFlow ● TensorFlow Lite ● Caffe Optimizations: ● Quantization ● Tensor Fusion MobileNet SSD SqueezeNet Inception Models: ● MobileNet SSD v1 & v2 ● SqueezeNet 1.0, SqueezeNet 1.1 ● Inception v1 & v2 & v3 3 & v4 Libraries: ● OpenCV ● OpenVINO QuantizationTensorRT Optimisation next input 3x3 CBR 3x3 CBR 1x1 CBR 1x1 CBR max pool next input input tensor Expansion layer Depthwise layer filter the data Projection layer uncompress the data compress the data outpt tensor Min Input (float) Min Max Quantize QuantizedRelu Dequantize Output (float) Ma x Eight Bit Min Ma x Eight Bit maxpool/2 conv1 fire2 fire3 fire4 fire5 fire6 fire7 fire8 fire9 fire10 softmax maxpool/2 globalavgpool maxpool/2 96 128 128 256 256 384 384 512 512 1000
  • 16. 16 Confidential 16 IFTTT Final State Machine ● Configurable FSM ● Generic data concept ● Visual simulation ● Visual monitoring States list Objects list Events list Action list
  • 17. 17 Confidential AIoT Edge and Cloud Accelerator Parts
  • 18. 18 Confidential 18 Edge AIoT Accelerator Module Battery AI Device Display Camera ● A fully standalone video processing device ● Does not require cloud or mobile device connection for operation presenting ● Contains replaceable parts ● Single hardware for any IoT solution demonstration ● Contains port for extended sensors and Ethernet with PoE ● ML models can be downloaded or updated from the cloud to support a specific IoT solution demonstration
  • 19. 19 Confidential 19 AIoT Accelerator Cloud Part Reaction Objects AI Camera Vision AB SM AI camera classifies objects ● Object attributes: ○ ID, probability, location, and state ● AB SM events: ○ Object ID change ○ Time of the day ○ Calendar ○ Public events ● AB SM reactions: ○ Change of state ○ Giving out notifications ○ Control of accelerators ○ Change Object attributes
  • 20. 20 Confidential 20 AIoT Accelerator Block Diagram Object Action Location Detection Event generation FSM updateActions Actions Internal ● Update Edge ML nodel ● Start record stream ● Take an image shoot ● Assign ID to object/location ● Change object attribute External ● Make a call ● Control switch/attenuator ● Send a sound message ● ETL action
  • 23. Confidential 23 LvM2M Data Model Upgrade ● Objects array ● Probability information ● Correlation data ● Neighbor sensor data
  • 24. Confidential 24 IFTTT STM Design ● Templates from different solution ● Custom Customization ● Visual simulation ● Visual history and action logic ● Ability for a third party device integration ● API for the external applications
  • 25. 25 Confidential Questions? Dov Nimratz Senior Solution Architect dov.nimratz@globallogic.com +380-91-333-3300
  • 27. 27 Confidential Thank you Dov Nimratz Senior Solution Architect dov.nimratz@globallogic.com +380-91-333-3300