Computer vision: British Clay
summary
Demo uploads 100 Wallace and 100 Gromit images to Google AutoML Vision, trains a small model in minutes (free tier under one hour), and reports ~85% precision / ~92% recall on validation while correctly classifying unseen photos at up to 99% confidence. Compares fast Microsoft Custom Vision (86% on a sample, mobile export) with Machine Box and Clarifai. Includes 201 "none of the above" negatives to cut false positives and emphasizes designing actionable labels.
AI-written summary · about
chapters (14)
- 0:00Introduction and setup
- 1:53Demo with AutoML Vision
- 6:11Machine learning options overview
- 11:36Transfer learning explained
- 13:52What is AutoML
- 17:11AutoML tools comparison
- 23:15Training with negative examples
- 24:44Google AutoML Vision results
- 29:11Precision vs recall trade-offs
- 31:39AutoML advantages summary
- 32:03Data cleaning importance
- 32:23Designing actionable labels
- 33:07Related AI talks today
- 34:29Q&A and closing
related talks
24mAutoML: No Code No Problem
57mBuilding smarter solutions with Machine Learning, from magic to reality
22mAutoML Vision: the PaaS of Machine Learning
47mBuilding smarter apps with Machine Learning, from magic to reality
33mComputer Vision, Not Just for Breakfast Anymore
01h 33mComputer Vision in Azure
44mBuilding smarter solutions with no expertise in machine learning
59mThe bold promise of teaching computers to understand the visual world