The document discusses Custom Vision Service, an AI tool that allows users to build custom image classifiers. It provides an overview of how Custom Vision Service works, including uploading images, training a model with labeled images, evaluating images with the trained model, and improving the model through active learning. Best practices for using Custom Vision Service like using a sufficient number and variety of images are covered. The document also discusses exporting Custom Vision models to containers and the edge, as well as Project Brainwave and using FPGAs to extend AI to the edge for increased processing power.
2. Cognitive Services
Computer Vision vs Custom Vision how they differ
Which scenarios to chose for either.
Time to Market
Taking it to the EDGE
Overview Project Brainwave and FPGA’s
Intel ComputeStick for edge devices
Agenda
5. What is it?
An API
What does it do ? Classification | Object Detection
What do I need to know ?
Object’s [] ModelCustom Vision
6. Custom Vision Service
Export to Container (export DockerFile + model +service) and to ONNX
S0 tier expanded to up to 250 tags and 50,000 images.
7. Upload Images
Upload your own labeled images, or use Custom Vision Service to quickly tag any unlabeled images.
Train
Use your labeled images to teach Custom Vision Service the concepts you want it to learn.
Evaluate
Use simple REST API calls to quickly tag images with your new custom computer vision model.
Active learning
Images evaluated through your custom vision model become part of a feedback loop you can use to keep
improving your classifier.
Custom Vision Service
8. Upload Images
Upload your own labeled images, or use Custom Vision Service to quickly tag any unlabeled images.
Train
Use your labeled images to teach Custom Vision Service the concepts you want it to learn.
Evaluate
Use simple REST API calls to quickly tag images with your new custom computer vision model.
Active learning
Images evaluated through your custom vision model become part of a feedback loop you can use to keep
improving your classifier.
Custom Vision Service
16. Best Practices for using Custom Vision
• Use at least 30 images for each tag
• Images should be the focus of the picture
• Use sufficiently diverse images and backgrounds (ex: cats
with red background and dogs with blue background)
• Train with images that are similar in {quality, resolution,
lighting, etc.} to the images that will be used in prod
• Supports Microsoft accounts (MSA) and AAD
17. Resources: Custom Vision Service
Get started at http://customvision.ai
Programmatic API access using C# (Python and Node
SDKs coming soon):
https://github.com/Microsoft/Cognitive-CustomVision-
Windows
18. Project BrainWave : FPGA’s
Get started at http://customvision.ai
Programmatic API access using C# (Python and Node
SDKs coming soon):
https://github.com/Microsoft/Cognitive-CustomVision-
Windows
21. Container Registry
Machine Learning
(Model Management)
Azure VM
Raspibian
Docker for
Linux
AML
Module
Picture
Jpeg File
Result
Serialized Data
Data Check
&
Determination
HTTP
Generate
Result SenseHat
Docker Apps
AML Package for
Computer Vision
Camera
Extending The AI to edge