Unit 2: Train Your Own Model¶
Goal: train a real, working machine learning model yourself, by example, using Google's free Teachable Machine — no coding required.
Privacy and training data
Webcam/microphone capture does not work through this lab's remote desktops (verified against the installed Guacamole version) — and separately, photos or audio of yourself or classmates are personally identifiable information, which raises real privacy/FERPA concerns in a school setting. For every project in this unit: use the upload existing files option with non-identifying subjects (objects, hand gestures framed to exclude faces, drawings), not live webcam capture of people. If an activity specifically calls for live camera testing, do that on your own personal device's browser, not through the lab portal.
Lessons¶
- Meet Teachable Machine
- Training an Image Classifier
- Testing and Iterating
- Confidence Scores and Uncertainty
- Exporting and Using Your Model
- Mini Project: Train a Classifier for a Real Task
Note
This guide is maintained by the instructors. If something is wrong or unclear, tell your instructor.