2. Hack
Your
Future
Machine Learning on Azure
Azure Cognitive Services
Computer Vision – Face API
Custom Vision
Text Analytics
Anomaly Detector – Metrics Advisor
3. Domain Specific Pretrained Models
To reduce time to market
Azure
Databricks
Machine
Learning VMs
Popular Frameworks
To build machine learning and deep learning solutions TensorFlow
PyTorch ONNX
Azure Machine Learning
Language
Speech
…
Decision
Vision
Productive Services
To empower data science and development teams
Powerful Hardware
To accelerate deep learning
Scikit-Learn
PyCharm Jupyter
Familiar Data Science Tools
To simplify model development Visual Studio Code Command line
CPU GPU FPGA
From the Intelligent Cloud to the Intelligent Edge
Azure Cognitive Services
4. Azure Cognitive Services
Language
Vision Decision
Speech
Recognize, identify, caption,
index, and moderate your
pictures, videos, and digital ink
content.
Convert spoken audio into text,
use voice for verification, or add
speaker recognition to your app.
Allow your apps to process natural
language with pre-built scripts,
evaluate sentiment and learn how
to recognize what users want.
Build apps that surface
recommendations for informed
and efficient decision-making.
Video Indexer
Computer Vision
Face
Custom Vision
Form Recognizer
Video Indexer
Speaker Recognition
Speech Translation
Speech to Text
Text to Speech
Language Understanding
Immersive Reader
Translator
QnA Maker
Text Analytics
Anomaly Detector
Content Moderator
Personalizer
Metrics Advisor
13. to introduce
drivers and riders right away
three weeks
That left us more time
to spend optimizing the user experi
Face API
14. “Now, companies are simply pushing sales people into the field and they’re
learning through experience— a ridiculously expensive way to train.
Every deal lost due to lack of confidence costs the company real money.
If we can minimize that and actually get sales people ready to sell,
it’ll have a huge impact on productivity,”
Jim Ninivaggi
Senior Vice President
Business Development
Problem
Training sales people
through experience is
ridiculously expensive.
Every deal lost due to lack
of confidence costs the
company real money.
Solution
Create a training platform that allows sales
reps to perfect their pitch through video and
Cognitive Services. Utilizing Face API,
Emotion API, and Text Analytics, we both
analyze their pitch, and feed an ML model to
provide feedback on their performance.
Power your Content. Power your Sales.
Brainshark
15.
16. Custom Vision API – a short explanation…
This service is an easy-to-use tool for prototyping, improving, and
deploying a custom image classifier to a cloud service, without any
background in computer vision or deep learning required.
“cat”
“dog”
Model
Custom Vision
17. Best Practices for using Custom Vision API
• Use at least 30 images for each tag
• Images should be the focus of the picture
• Use sufficiently diverse images and backgrounds.
• Train with images that are similar in {quality, resolution, lighting, etc.} to the images that will be
used in prod
• Current project limitations while in preview: 1000 images, 50 tags, 20 iterations saved
• Current account limitations while in preview: 20 projects, 1000 predictions per day
18. Example Customer Scenarios
Customer Support
• Enable a customer to identify a product for support by taking a photo. No finding the manual or pulling the
appliance out to identify it!
Service Engineers
• Identify parts for ordering
Manufacturing
• Fault detection on assembly lines to avoid machine downtime and drop in production rates (provided
differences are obvious)
Data Scientists
• Automatic tagging instead of manual, to create features or labels
25. Recognition of unusual patterns of behavior in data that don’t conform to expected outcomes.
What is Anomaly Detection
26. The Challenges from Real-World Data
Manual rule
setting won’t
scale and adapt
AD learns from the data on rules
which differentiate outliers from
normal pattern automatically
AD automatically selects the best
pre-built model from model pool
behind the scenes
AD hides the complexity and
provides ONE intuitive parameter
to change sensitivity
Many types of time
series that no single
algorithm fits all
Many existing solutions
require data science
knowledge
27. Anomaly Detector Service
Take time series
as input
Auto model selection and
inference
Return anomaly related
metadata (is Anomaly,
range of expected value…)
Microsoft Cognitive Services
To ensure the health of your business, you want to track
your key metrics like revenue and understand whether
something is out of historical pattern.
Sensor time series data, you want to be alerted on the
drifting which could imply system malfunctions.
Example: Example:
28. Metrics Advisor (preview):
Metrics Advisor is a part of Azure Cognitive Services that uses AI perform data monitoring and anomaly detection in
time series data. The service automates the process of applying models to your data, and provides a set of APIs web-
based workspace for data ingestion, anomaly detection, and diagnostics - without needing to know machine learning.
Example: Example:
30. When Should I Use Anomaly Detector?
Real-Time
Business Health
You have KPIs reflecting business
& product health, you want to
monitor them 24X7 to avoid
business loss
Interactive Data Analytics
Analyzing metric data to
understand whether the data
contain anomalies out of
historical pattern
IoT – Remote
Monitoring
Monitor status of a system or
device and get early warning of
potential anomalies
31. Benefits
Easy to setup and
configure
Automatically ingest all of your data
feeds with a single API.
Powerful inference
engine
Intelligently applies multiple
anomaly detection approaches
Completely self-
training
No need for lots and lots of labeled
training data.
Offers customization
capabilities
Tune sensitivity of anomaly
detection based on your needs.
Scales to large
volumes
Ingest large amounts of data without
worrying about performance.
Works in Cloud and
Edge
Consume from the Cloud API or
deploy on edge devices as container.
32. Proven Technology
within Microsoft
• 400+ teams across Azure,
Windows, Office, Bing…
• Millions of time series
• Thousands of active users
within Microsoft
33. Basic SKU(Stock-Keeping-Unit) Case Study:
Data cleansing for predictive maintenance
• Using aircraft data (potential & available data) to generate predictive maintenance models to:
– Increase fleet availability
– improve mission & material planning
– reduce maintenance burden
• Challenges:
Equipment
interferences
Arc in
micro-secs
Military
Data
Noisy
Data
35. Anomaly Detector
“Innovation has always been a driving
force at Airbus. Using Anomaly
Detector, an Azure Cognitive Service,
we can solve some aircraft predictive
maintenance use cases more easily.”
Peter Weckesser
Digital Transformation Officer
Airbus
Anomaly Detector video
36. Situation: Solution: Impact:
“Only Microsoft offered the proven, cutting-edge technologies and models that have
been trained on Microsoft data sets on a very large scale and are available completely
disconnected.... All of this accelerates our time to market, and that has been the key
differentiator for us.”
Airbus innovates continually, ever
mindful of the strict security
requirements that constrain many of its
customers, like government agencies
and international security
organizations. It also envisioned AI
applications to reimagine answers to
complex problems.
—Marcel Rummens, Product Owner of Internal AI Platform, Airbus
The company created its own
restricted cloud with Azure AI
solutions, like its aircraft anomaly
detector. It used Azure Cognitive
Services to create a pilot training
chatbot and a predictive
maintenance solution based on
Anomaly Detector.
Airbus uses the built-in functionality of
Cognitive Services and Azure AI solutions
to hasten development, shortening time to
market. The solutions it’s creating have
myriad benefits, from optimizing military
aircraft maintenance to making pilot
training more effective.
Customer:
Airbus
Industry:
Defense and Intelligence
Size:
10,000+ employees
Country:
Germany
Products and services:
Microsoft Azure
Microsoft Azure AI
Microsoft Azure Anomaly Detector
Microsoft Azure Cognitive Services
Microsoft Azure Cognitive Speech to Text
Microsoft Azure Cognitive Text to Speech
Microsoft Azure Kubernetes Service
Read full story here