Build an AI-powered Pet Detector in Visual Studio Code
summary
Builds a pet classifier inside Visual Studio Code using Jupyter notebooks, TensorFlow and MobileNet transfer learning, trains on the Oxford Pet dataset (37 classes, ~200 images per class), gets ~78% after 500 local steps in ~8 minutes, then runs hyperparameter search with Azure Hyperdrive 200 runs -> ~90% using early termination, converts the notebook into a Python module, registers and deploys the model as a cloud endpoint that returns JSON predictions, and the demo predictions work end to end.
AI-written summary · about
chapters (15)
- 0:00Building a pet detector
- 1:18Live demo
- 2:48Human vs machine pet classification
- 4:55Deep vs traditional ML
- 7:21Traditional ML workflow
- 10:07Oxford pet dataset
- 13:13Training with limited data
- 14:59Transfer learning MobileNet
- 17:26Hyperparameter tuning
- 23:19Cloud production
- 27:12Workflow review
- 29:43Q&A data validation
- 31:07Q&A transfer vs k-fold
- 32:10Q&A transfer learning limits
- 33:15Q&A VS Code Pygame
related talks
11mAll about AI Toolkit for Visual Studio Code
57mEasier AI allows everyone to build smart apps
24mDon't want to miss out on the AI wave?
25mBuilding AI Agents
01h 00m5 Levels of AI in Microsoft Azure
32m"How Smart is AI? – Real projects to build real understanding"
44mAI for every developer
01h 00mAI for Every Developer