Keynote: From Zero to ML on Google Cloud Platform
summary
Shows concrete workflows on Google Cloud Platform: use Cloud Vision and Natural Language APIs for one-call predictions, customize with Auto ml vision (alpha whitelist only) requiring about 100 images recommended 10 minimum for a prototype, or build from scratch with TensorFlow + Cloud ML Engine. Demonstrates transfer learning to ship a Taylor Swift detector trained on ~200 images via the TensorFlow Object Detection API (models return bounding boxes and ~99% confidence), and a county voting classifier over ~3,000 counties reaching ~96–97% accuracy.
AI-written summary · about
chapters (14)
- 0:00Intro
- 3:46What is machine learning
- 7:18How to use machine learning
- 14:48Vision API demo
- 17:18Natural Language API
- 19:37Demo
- 21:21Recap
- 23:03AutoML Vision
- 30:32Custom Prediction Tasks
- 31:44Rolling Your Own Model
- 34:35Cloud ML Engine
- 39:35Custom ML Task
- 45:29Running the Model
- 48:41Auto ML Vision
related talks
44mBuilding smarter solutions with no expertise in machine learning
27mKeynote - Google Cloud, Machine Learning
57mBuilding smarter solutions with Machine Learning, from magic to reality
54mBuilding smarter apps with Machine Learning, from magic to reality
47mBuilding smarter apps with Machine Learning, from magic to reality
42mMachine Learning APIs, from Training [...]
42mUse Machine Learning in your code, without being a ML expert
59mMachine Intelligence at Google Scale