Intel IoT Edge Computing 在 AI 領域的應用與商機
- 1. Intel IoT Edge Computing
在 AI 領域的應用與商機
• James Shih
• Director of Platform Application
• IoTG, INTEL
- 2. Internet of Things Group
1. IDC
2. MC/EDC: The Digital Universe of Opportunities
3. Goldman Sachs
4. IMS Research
2
Things network cloud
50B
devices1
44
zetabytes2
212B
sensors1
85%
unconnected4
Cost of Sensors1
Past 10 Years
Cost of Bandwidth2
Past 10 Years
Cost of Processing3
Past 10 Years2X 40X 60X
- 3. Internet of Things Group
InternetofThings-ThreephasesofIoTdeployment
CONNECTED SMART AUTONOMOUS
Connect the Unconnected Smart & Connected Things Devices Communicate Amongst
Themselves at Local
IoT Gateway is Fundamental in Moving Data for IoTEdge Computing is Fundamental for IoT
- 8. Intel® CONFIDENTIAL – all material covered by NDA
Vision is Driving Artificial Intelligence at the Edge
2500petabytes Daily
videosurveillancedata
generatedin2019
- 9. Intel® CONFIDENTIAL – all material covered by NDA 9
Deep learning revenue
to grow from $655
Million in 2016 to $35
Billion by 2025
--a CAGR of 53%2
The overall computer
vision market to reach
$17.38 Billion by 20231
The video analytics
market to reach $11.17
Billion by 2022 at a
CAGR of 21.5%3
Bandwidth & Storage Costs
Near Real Time Response Needs
Forensic to Predictive
Accuracy
Privacy
DeepLearning
1. Marketsandmarkets, Computer Vision Market, 2017
2. Tractica, 2Q, 2017
3. Marketsandmarkets, Video Analytics Market, 2017
- 10. Intel® CONFIDENTIAL – all material covered by NDA
Industrial SmartCities Healthcare
Retail
10
Fast & Accurate Defect
Detection
Identify Congestion &
Redirect Traffic
Interpret Shopper Behavior &
Improve Inventory Tracking
Diagnose Illnesses with Real-
time Feedback
Behavior Anomalies &
Property Demographics
DigitalSecurity&Surveillance
- 11. Internet of Things Group
If need higher
DL performance
utilize:
Openvino™Toolkit
Writeonce - deploytoCPU,GPU,FPGA&VPUAccelerators - LeveragecommonLibraries/Frameworks/networks
Integrated
Graphics
(on die with Intel CPU)
Accelerators
(on board or add-in cards)
VPU FPGA
Host CPU
App & Media Processing,
IO, Storage
If need higher
performance
utilize:
- 12. DEEP LEARNING
Inference
Engine
Open CV
CV Library
(Kernel & Graph APIs)
Open CL
>100 Networks
Supported
CV
Algorithmes
COMPUTER VISION
ONNX
HIGH PERFORMANCE & SCALABILTY FOR THE EDGE
AGNOSTIC, COMPLEMENTARY TO MAJOR FRAMEWORKS
CROSS-PLATFORM FLEXIBILITY
Model
Optimizer
*Other names and brands names may be claimed as the property of others
- 13. Intel® CONFIDENTIAL – all material covered by NDA
Load, infer
GPU = Intel CPU with integrated graphics processing unit/Intel® Processor Graphics
OpenCL and the OpenCL logo are trademarks of Apple Inc. used by permission by Khronos
*Other names and brands may be claimed as the property of others.
Caffe*
TensorFlow*
MxNet*
IR = Intermediate
Representation
format
Convert &
optimize to fit all
targets
CPUPlugin
GPUPlugin
FPGAPlugin
MyriadPlugin
Model
Optimizer
Convert &
Optimize
Extendibility
C++
Extendibility
OpenCL™
Extendibility
OpenCL/TBD
Extendibility
TBD
Trained
Model
Inference
Engine
Common API (C++)
Optimized cross-
platform inference
Model Optimizer
What it is: Preparation step -> imports trained models
Why important: Optimizes for performance/space with
conservative topology transformations; biggest boost is
from conversion to data types matching hardware.
Inference Engine
What it is: High-level inference API
Why important: Interface is implemented as dynamically
loaded plugins for each hardware type. Delivers best
performance for each type without requiring users to
implement and maintain multiple code pathways.
IR .data
IR
13
- 16. 2018INTELVISION PORTFOLIOLaunch
STRONG INDUSTRY PARTNERSHIP
Other names and brands may be claimed as the property of others. Sale Pipeline as of October
INTEL®VISION
ACCELeRATORDESIGN
INTEL®VISION
ACCELERATORDESIGN
With Intel® Movidius™ VPU
1X→8XVPU
With Intel® Arria™ 10
FPGA
Writeonce | Leveragecommonalgorithms
deploytoCPU,GPUandAIAccelerators
- 18. Intel® CONFIDENTIAL – all material covered by NDA
Usbvisiondevkits Intel® Movidius™ Neural Compute Stick + Intel® Movidius™ Myriad™ 2 VPU
VisionAccelerator
Designproducts
Vision
Accelerator
Kits
Intel® Core® Processor + Intel® Vision Accelerator
Design with Intel® Arria® 10 FPGAbetter
good
edge & cloud
developers, OEMs, ISVs
cloud developers
TargetAudience
OEMs
Integrators, cloud
developers, edge
developers
developers
end users
Intel®iotMarketReadySolutions Smart Cities Retail Industrial
Worker Safety
Smart
Cities
Transportation
Visionkits| CSPintegration AWS GreenGrass* Inference
Implementation
RFPReadykits Health Robotics Retail
best
Security and Surveillance
Intel® Atom™ Processor + Intel® Vision Accelerator
Design with Intel® Movidius™ Myriad™ 2 VPU
Intel® Atom™ Processor + Intel® Vision Accelerator
Design with Intel® Movidius™ Myriad™ X VPU
Intel® Core® Processor + Intel® Vision Accelerator
Design with Intel® Movidius™ Myriad™ X VPU
Intel® Xeon® Processor + Intel® Vision Accelerator
Design products
Intel® Vision Accelerator Design with Intel®
Arria® 10 FPGA
Intel® Vision Accelerator Design with Intel® Movidius™
Myriad™ 2 VPU
- 19. Intel® CONFIDENTIAL – all material covered by NDA
VisionKit
poweredbyIEITANK
VisionKit
poweredbyAAEONUP²
Developer Kits to Accelerate Computer Vision Deployments article
*Other names and brands may be claimed as the property of others.
Intel®distributionof
OpenVino™Toolkit
Download Free
Learn More
Get started quickly with:
Developer resources
Intel® Tech.Decoded online webinars, tool
how-tos & quick tips
Hands-on, in-person workshops
Support
Connect with Intel engineers & computer
vision experts at the Community Forum
UP Squared* Board (Intel Atom® x7-E3950)
AI Core- Intel® Movidius Myriad 2 mini PCIe*
VPU
USB camera
Intel Distribution of OpenVINO toolkit
Ubuntu* 16.04 Desktop w/ 4.14 kernel
Intel® System Studio
Arduino Create*
IEI TANK 870-Q170
(6th/7th Gen Intel® Core™ processor)
Intel Distribution of OpenVINO toolkit
Intel® Media SDK
Ubuntu Desktop16.04 Desktop LTS
Intel System Studio
link to purchase link to purchase
Futureproof
design
Expeditepath
toproduction
Accelerate
performance
19Optimization Notice
- 20. Intel® CONFIDENTIAL – all material covered by NDA 20
Rapid Prototyping, Validation, and Deployment
of Deep Neural Network Inference at the Edge
T R A I N O P T I M I Z E D E P L O Y
Mode
l
+
+
+
with Intel® Vision
Accelerator Designs
featuring Intel®
Movidius™ Myriad™ X VPU
P R O D U C T I Z E
your deep neural
network on the Intel®
Neural Compute Stick 2
and your prototyping
host
your pre-trained models with
the Intel® Distribution of
OpenVINO™ toolkit
your deep neural
network models where
desired
Write once, deploy everywhere with the
Intel® Distribution of OpenVINO™ toolkit
D E V E L O P A N D S C A L E
through Intel® AI: In Production
- 21. Intel® CONFIDENTIAL – all material covered by NDA
*Other names and brands may be claimed as the property of others.
Intel® Movidius™
VPU Hardware
TensorFlow*, MobileNet
SSD* and YOLO* v2
Worker Safety Management
21
Intel® Distribution of
OpenVINO™ Toolkit
MobileNet SSD*
YOLO v2*
Intel® Arria® 10
FPGA Hardware
Caffe*,
MobileNet SSD*
and CL-DNN
Technology
Medical Image Detection
Intel® Distribution of
OpenVINO™ Toolkit
MobileNet SSD*
CL-DNN*
Optimization Notice
- 22. 22Intel Confidential
Whatis5G?
• Next generation of wireless networks
• Will provide higher speeds, greater capacity, and lower latency
• Will be capable of supporting billions of connected devices and ‘things’
• Requires fundamental transformation of the network infrastructure
• Extends intelligence throughout the network
Reactive, smart,
and connected
devices
Faster data rates
Data and the
‘app’ revolution
Cellular Comms.
5G
4G
2G
3G