Mais conteúdo relacionado Semelhante a NEW LAUNCH! AWS DeepLens workshop: Building Computer Vision Applications - MCL212 - re:Invent 2017 (20) Mais de Amazon Web Services (20) NEW LAUNCH! AWS DeepLens workshop: Building Computer Vision Applications - MCL212 - re:Invent 20171. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
AWS re:INVENT
AWS DEEPLENS
B u i l d d e e p - l e a r n i n g p o w e r e d c o m p u t e r v i s i o n s k i l l s
November 29, 2017
Jyothi Nookula, Senior PTM, DeepCam
Sunil Mallya, Sr. AI Solutions Architect
Aran Khanna, AI Software Dev Engineer
MCL212
2. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
JYOTHI NOOKULA SUNIL MALLYA
MEET THE TEAM
ARAN KHANNA
3. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
AWS DEEPLENS
IS NOT A
VIDEO CAMERA…
…IT’S THE
WORLDS FIRST
DEEP LEARNING
ENABLED
DEVELOPER KIT
4. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
GET STARTED WITH SAMPLE PROJECTS
ADD CUSTOM FUCTIONALITY
OR
CREATE YOUR OWN PROJECT
ARTISTIC STYLE
TRANSFER
OBJECT
DETECTION
FACE DETECTION /
RECOGNITION
HOT DOG / NOT
HOT DOG
CAT VS. DOG
ACTIVITY
DETECTION
5. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #2: DEPOLYING
A PROJECT
LAB #1: ML OVERVIEW LAB #4:
EXTENDING A
PROJECT
TODAY WE WILL COVER
LAB #3: MODEL
EDITING
6. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #1
M A C H I N E L E A R N I N G O V E R V I E W
7. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #1: FOUNDATIONS OF ML
• Objective: Learn key elements of ML and how DeepLens works
• Time: 10 mins
• Steps:
1. Machine learning 2. About DeepLens
8. AI, ML, and DL
Hardware and
learning
algorithms Fit a function
to the data
Fit a network
structure to the data
9. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Model training Inference
OVERVIEW OF DEEP LEARNING
Data
10. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
DATA
Annotate Preprocess Data split
11. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
MODEL TRAINING
• Define model architecture
• Input the annotated and cleaned data into the model
• Multiple iterations (epochs) to train the model
• Validate with held back dataset
Large,
annotated
dataset
Training set
Validation set
Training
Validate
12. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
INFERENCE
It’s where the magic happens!
1. Preprocess new data/image just like training set.
2. Feed image back to the trained model to get a predicted output.
13. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
AWS DEEPLENS ARCHITECTURE
Video out
Data out
I N F E R E N C E
D E P L O Y P R O J E C T S
Manage device
Security
Console Project
Management
AWS Cloud
Intel: Model Optimizer
cIDNN and Driver
14. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
DEEPLENS SPECIFICATIONS
• Intel Atom Processor
• Gen9 graphics
• Ubuntu OS- 16.04 LTS
• 100 GFLOPS performance
• Dual band Wi-Fi
• 8 GB RAM
• 16 GB Storage (eMMC)
• 32 GB SD card
• 4 MP camera with MJPEG
• H.264 encoding at 1080p resolution
• 2 USB ports
• Micro HDMI
• Audio out
• AWS Greengrass preconfigured
• clDNN Optimized for MXNet
15. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
QUESTIONS?
16. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #2
D E P L O Y I N G A P R O J E C T
17. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #2: HOW TO DEPLOY A PROJECT
• Objective: Get started with a sample project
• Time: 30 mins
• STEPS:
1. FIND AND
SELECT A
PROJECT
2. CONFIGURE
3. DEPLOY &
TEST
4. HOW IT
WORKS
18. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
FIVE MINS TO YOUR FIRST CV PROJECT
LET’S GET STARTED!
WAKE UP YOUR
AWS DEEPLENS
(wiggle the mouse)
19. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LOG IN
Password: Aws2017!
20. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
ACCESS DEEPLENS CONSOLE
1. Open QuickLinks File from the desktop.
2. Click Quick Link #1 labeled DeepLens Console.
(console.aws.amazon.com/deeplens)
21. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LOGIN?
If the AWS DeepLens Console login was not saved …
1. Check the sticker on the side of your DeepLens.
2. Enter you account ID, username, and password.
22. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
CREATE A PROJECT
• Log in to Console. Use the username
and password on the sticker on the
side of your device.
• This is the AWS DeepLens Console.
• Click the Create Project button.
23. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
USE OBJECT DETECTION SAMPLE
1. Select Use a project
template.
2. Select Object Detection
from the sample project
templates.
3. Select Next at the bottom
of screen.
24. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
NAME PROJECT
1. Edit project name
(e.g., Object-
detection-MikeM).
2. Optional description
(leave blank).
3. Click Create.
25. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
DEPLOY PROJECT TO THE DEVICE
1. Find your project in
the list (the one you
just named).
2. Select the radio
button.
3. Click Deploy to
device.
26. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
1. Choose your device.
2. Click Review.
TARGET YOUR DEVICE
27. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
DEPLOY!
• Click Deploy.
• A note on costs …
28. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
WAIT FOR PROJECT TO BE DEPLOYED
Blue banner = Deployment in
progress
Green banner = Deployment
successful
29. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
WHILE WE WAIT…
• Cloud to device
• On the device
30. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
UNDER THE COVERS - CONSOLE
31. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
UNDER THE COVERS – DEVICE
32. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
For the Workshop:
1. Open QuickLinks
(existing tab)
2. Cut & paste “mplayer”
string into Terminal (Live
or Project Stream)
Console has instructions:
VIEW OUTPUT
33. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
QUESTIONS?
34. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #3
M O D E L E D I T I N G
35. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #3: MODEL EDITING
• Objective: You will learn how can use the existing object detection
model to create a hot dog detector using AWS SageMaker
• Time: 40 mins
• Steps:
1. ABOUT AWS
SAGEMAKER
2. SELF-PACED LAB
36. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
End-to-end
machine learning
platform
Zero setup Flexible model
training
Pay by the
second
INTRODUCING AMAZON SAGEMAKER
The quickest and easiest way to get ML models from idea to production
$
37. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB DETAILS
The self-paced lab will include the following steps:
1. Import a prepared Jupyter Notebook into SageMaker.
2. The notebook walks through modifications to the squeezenet
object dection model, to turn it into a binary classifier.
3. Create an S3 bucket and export the updated model there.
4. Import model to a new DeepLens Project.
5. Deploy & test!
6. If time: Deploy the Sample Project ‘Hot dog recognition’.
38. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #3 – SELF PACED
1. Use QuickLinks file (open in Firefox).
2. Click on the GitHub link for the Lab 3 instructions!
40 minutes
39. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
TIME’S UP!
QUESTIONS?
40. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #4
E X T E N D I N G A P R O J E C T V I A L A M B D A
41. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #4: EXTENDING A PROJECT VIA LAMBDA
• Objective: Be able to extend the functionality of a project.
• Time: 15 mins
• Steps:
1. ABOUT LAMBDA
FUNCTIONS ON
DEEPLENS
2. SELF-PACED LAB
42. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
ABOUT LAMBDA’S ON DEEPLENS
43. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
44. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB DETAILS
The self-paced lab includes the following steps, all performed in
the cloud.
Assumption: Default Hot dog recognition Project publishes model
results to an IoT Topic in the cloud via MQTT.
1. Create a Lambda function.
2. Create an IoT rule.
3. Integrate the IoT rule with AWS DeepLens.
4. Trigger the Lambda function when a hotdog is detected.
45. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
LAB #4 – SELF PACED
1. Use QuickLinks file (open in Firefox).
2. Click the GitHub link for the Lab 4 instructions!
15 minutes
46. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
TIME’S UP!
QUESTIONS?
47. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
DREAM. LEARN. BUILD. WIN!
Announcing the AWS DeepLens Challenge Virtual Hackathon
Nov 29th – Jan 24th
$7,500 USD
2 tickets to re:Invent 2018
Special edition custom DeepLens
$5,000 USD
1 ticket to re:Invent 2018
Special edition custom DeepLens
$2,500 USD
Special edition custom DeepLens
PARTICIPANT PRIZE:
EVERY ELIGIBLE SUBMISSION GETS A CUSTOM SKIN FOR THEIR DEEPLENS WORTH $34.99 MSRP
SIGN UP NOW:
awsdeeplens.devpost.com
48. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
WHAT WILL YOU BUILD WITH
DEEPCAM?
49. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
THANK YOU!