Explore image filters, train a small CNN, inspect forward and backward calculations, play with hand tracking, and build a Python video effect. The live demo runs in the browser. Camera processing stays on the device.
Exercises → Lecture → Playground → Go deeper → Final challenge → Take-home.
The first page is start.html. The repository root index.html is the final rock–paper–scissors game; use the workshop link above to begin with the exercises.
Run python serve.py on Windows or python3 serve.py on Mac. Open the URL printed by the launcher. Keep the complete folder together. The browser demo needs no pip/npm installation. The optional Python take-home renderer has its own requirements file.
For camera use, allow browser camera access over HTTPS or localhost. Mouse, touch, generated-image and illustrated practice activities remain available without a camera.
Python examples are in python/ and take-home-code/. Preserve vendor license notices, photo provenance in references/photo-bank-provenance.json, and video source/license links in the take-home pages. Bundled pretrained models are third-party inference assets; the symbol CNN is trained locally in the browser.