Beyond the Usual Suspects: Emerging Cognitive Services
summary
A survey of lesser-known Microsoft Cognitive Services showing concrete demos and results: Content Moderator for PII detection and profanity filtering (~100 languages, ~$0.84/1,000 transactions); Computer Vision plus Celebrity/Landmark APIs for bounding boxes, labels and face counts; Custom Vision export to Docker to run trained models offline on Raspberry Pi and mobile from dozens to thousands of images; real-time Speech Translation streaming audio-to-text and translating with ~10–15s latency; Speaker Authentication using three enrollment phrases for yes/no verification; Project Prague for local Kinect-based gesture control; and Project Cuzco for extracting and clustering web event signals.
AI-written summary · about
chapters (12)
- 0:00Emerging cognitive services intro
- 2:33Core cognitive services
- 9:05Computer vision vs content moderation
- 14:07Custom vision training
- 20:34Speech translation
- 27:33Speaker recognition auth
- 31:06Microsoft AI labs experiments
- 34:24Event tracking anomalies
- 37:34AI search Q&A
- 39:17Gesture recognition demos
- 44:13Quick prototyping
- 46:02Q&A
related talks
01h 05mMicrosoft Cognitive Services: Making AI Easy
35mExplore Cognitive Services
01h 00mA lap around Microsoft Cognitive Services Vision API
01h 03mEasier AI allows everyone to build smart apps
46mBuilding accessible experiences with AI
49mMake your applications intelligent (...)
57mEasier AI allows everyone to build smart apps
51mKeyboards? Where we’re going, we don’t need keyboards
