A lap around Microsoft Cognitive Services Vision API

on YouTube ↗01h 00m

" summary

Live demos show Microsoft Cognitive Services Vision APIs as REST endpoints that identify faces, emotions, objects and text, and scale to production. Custom Vision classifies objects trained with 40 images and hits ~99% confident on Coke vs Pepsi in poor light; a PUV ultrasound classifier uses ~21 examples and flags anomalies at ~99% confidence. Face API returns age/gender and matches identities in milliseconds. Video Indexer extracts names/keywords; CNTK is available for building custom deep models. Pricing: free to prototype, cents per call in production.

AI-written summary · about

" chapters (16)

1. 0:00Start
2. 6:11Background
3. 12:24Read Out Text
4. 14:22Tools for Computer Vision
5. 16:21Demo
6. 29:36Intelligent Kiosk
7. 31:09Computer Vision
8. 32:37Categorize or Tag Images
9. 33:42Custom Vision Demo
10. 34:12Vision Api
11. 35:55Custom Vision Portal
12. 52:50I Gave You some Use Cases like another Creative Use Case I Came I Ran into the Other Day Was this Brilliant Guy Who's Using Predictive and Nano Analytics To Predict the Winner of Horse Races or Gambling He's Using Machine Learning for Gambling It's the Perfect Job Right You Have no Customers and You Have no Employees Right All Right Ok Let's Wrap this Up Oh One of the Last Things I Want To Talk about Very Briefly Is the Content Moderator this Is Not an Api this Is You Could Get the Source Code to this Thing this Is a Full-Blown Application that Microsoft Provides to a Number of Companies To Do Content Moderation
13. 54:17And It Will Pull the Metadata out of the Video and Go Farther than that It Will Do Machine Intelligence on the Video Meaning if this Video We'Re Recording Today Was To Run through this Tool It Would Automatically Pick Up My Name It Would Automatically Pick Up that We Talked about Machine Learning and Microsoft Cognitive Custom Vision It Would It Would Nail Down a Bunch of the Use Cases That We Talked about It Indexes Keywords so You Could Go to the Place in the Video like for Instance for the Build Videos
14. 54:53You Know I Run Them through this because I Know at some Point They Want To Type They'Re Talking about a Technology That I'M Interested in and I Could Just Zoom Forward to that I Don't Have To Go Through like They You Know in the Beginning of this Presentation that the Five Minutes Uncomfortable about Tim Slides I Could Just Skip that and Go Right to Custom Vision Which Would Be a Good Example I Believe How Are We Doing on Time Yeah I Got Four Minutes So Let Me Just Show You this Thing Real Quickly Hopefully Powerpoint Doesn't Freak Us Out Again Yeah Oh Perfect So I'M Logged In under My Account
15. 57:01This Is How You Get Started It's all for Free if You Want You Can Take a Picture of this or You Can Send Me an Email I Won't Save Your Email Address I Won't Put You on a List I Don't Even Know How To Do that but I Just Don't Have a Download Place for You yet Where You Can Download You Know Not Only the Powerpoint Deck I Have but the Demos That I Showed You and Where To Get those so that Would Be Here Top Right if You Send Me an Email and Want this Deck or Anything Else Just Send Me a Note I Promise You I Won't Keep Your Email Address
16. 58:51If You Want To Run this in Your Own Data Center or on Your Own Camera Stuff like It's GonNa Be More Expensive Just Assume that but Yeah in the Roadmap for All this Stuff Great Question L eles Question Was What's the Level of Accuracy for the Video Indexer You Might As Well Ask What's the Level of Accuracy for a Demographic Profile for Facial Recognition for Object Recognition So Then I Can Answer It Depends or I Don't Know I Don't Know Offhand I Could Tell You that in Video Indexer It's Shockingly Accurate like Whoa How Did You Do that because All the Natural Language Processing That that Problems Been Solved and Basically It's Looking for What's Being Spoken

" related talks

• 01h 05mAI for the Masses - Microsoft Cognitive ServicesStephen Bohlen · TechBash 2018
• 01h 05mMicrosoft Cognitive Services: Making AI EasyJennifer Marsman · NDC London 2017
• 44mAdding Image and Voice Intelligence to Your AppsDavid Giard · PyData Chicago 2016
• 46mBuilding computer vision applications Q&AShawndell Elfring, Kellen Anderson, Courtney Brewer, Kate Browne, Tonye Iyalla, Fisayo Okikiolu, Cenyu Zhang · Microsoft Build 2023
• 47mBeyond the Usual Suspects: Emerging Cognitive ServicesHank Boelman · NDC Oslo 2018
• 01h 03mEasier AI allows everyone to build smart appsDavid Burela, Azadeh Khojandi · NDC London 2018
• 57mEasier AI allows everyone to build smart appsDavid Burela, Azadeh Khojandi · NDC Sydney 2018
• 46mBuilding accessible experiences with AIHenk Boelman · Build Stuff 2021