The bold promise of teaching computers to understand the visual world
summary
Company builds the election-night TV map used by major networks and dozens of state voting apps, then demonstrates real systems using Microsoft Cognitive Services and the Azure REST APIs from a $20 camera and a Surface Book. A custom object classifier (called custom vision) trains Coke vs Pepsi from ~40/28 images in ~20 minutes and returns 99%+ scores; another model flags a rare fetal bladder disorder from 40 normal / 21 positive examples. Retail demos claim a 25% sales uplift and surveillance-scale deployments use thousands of machines.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 4:26Machine Learning
- 8:22The Journey
- 13:37The US Goes
- 15:55Moores Law
- 18:17Implications of Moores Law
- 20:56Machine Learning isnt a bunch of robots
- 21:37Competition is good
- 22:21Use case
- 25:17Proof of concept
- 28:36PowerPoint crashes
- 37:15Demo at home
- 38:40The vision of objects
- 48:15Conclusion
- 51:55Hololens
related talks
33mBuilding end-to-end computer vision solutions from deep learning models
58mPlaying the game with Azure Cognitive Services
27mLive and Machine Learn
42mProduct Recognition in Retail/E-Commerce with Azure AI
40mMachines are Learning
33mApplied AI: Real-world use cases for Microsoft's Azure Cognitive Services
01h 01mGetting Started with Computer Vision
42mProduct Recognition in Retail/E-Commerce with Azure AI