Building and Training your own Custom Image Recognition AI
summary
Uses Azure Cognitive Services with simple REST calls to build recognizers: Face API for person groups and Custom Vision for arbitrary objects. Demo creates a person group, uploads ~30 photos for three people and trains a model that identifies faces at ~81–82% confidence. A product classifier trains on ~60 images and predicts a new photo as product6 at 98%. A car model scrapes Bing to collect ~50 makes × ~150 images (requires manual cleanup), deploys as a prediction API and can export to TensorFlow/CoreML for offline use.
AI-written summary · about
chapters (18)
- 0:00Start
- 1:23Cognitive Services
- 4:52Create a Person Group
- 10:29Computer Vision Api
- 13:02Api Reference
- 14:27Ways To Send an Image
- 16:23Manage a Person Group
- 30:45Custom Vision Service
- 33:25Create a Brand New Project
- 33:55Add Images
- 35:49Test Page
- 40:32Prediction Url
- 44:49Output
- 47:46Web Services
- 52:01Image Detection
- 56:10Api for Recognizing Emotion
- 57:27Writing Your Own Machine Learning Algorithm
- 59:20Cognitive Services That Do Work with Videos
related talks
42mProduct Recognition in Retail/E-Commerce with Azure AI
42mProduct Recognition in Retail/E-Commerce with Azure AI
57mEasier AI allows everyone to build smart apps
23mBuild a look alike engine with machine learning and Angular
01h 09mBuild your own KITT with Azure Cognitive Services
16mHow to get started with Computer Vision AutoML in Azure
58mPlaying the game with Azure Cognitive Services
01h 00mA lap around Microsoft Cognitive Services Vision API