Deeper into ARKit with CoreML and Turi Create
summary
An app detects real bills with ARKit image tracking, captures a selfie, runs AVFoundation face detection and Firebase ML Kit contouring, then applies Core ML style transfer models produced with Turi Create hidden parameters. Initial training takes five days on a laptop but drops to ~2 hours per model on GPU; the final Core ML file is ~6.9 MB. About 180 models tried reveal which loss multipliers, input sizes and fine‑tuning choices actually produce usable face‑on‑bill results.
AI-written summary · about
chapters (6)
- 0:00Start
- 4:58Machine Learning
- 7:15Face Contour Detection
- 9:02Artistic Style Transfer
- 11:20Preparing Data
- 15:44Training
related talks
22mGetting Started with Apple ML
40mNavigating the Machine Learning Landscape for Apple Devices
30mLearning CoreML with fun
30mMachine Learning in Android Apps with ML Kit
29mFirebase MLKit — AI in Android Town!!
29mPushing Machine Learning into your app When TensorFlow and MLKit rule the world
22mMemojis under the hood
21mGetting the best of on-device ML with ML Kit