From TensorFlow to ML Kit: power your Android application with ML
summary
A hands‑on timeline: an app called Magritte is built to recognize daily objects and speak their names using transfer learning and a custom classifier, starting with an 80 MB TensorFlow model that infers in ~3–4 s. Quantization shrinks the model ~75%, then switching to MobileNet cuts CPU and memory sharply and makes inference realtime. Conversion to TensorFlow Lite and hosting models via ML Kit/Firebase enables OTA updates (no download callbacks yet). A toy robot with on‑device inference works but misclassifies under bad lighting.
AI-written summary · about
chapters (8)
- 0:00Introduction to edge ML
- 2:56First steps with models
- 5:05Neural network basics
- 9:10Training and optimization
- 12:58Mobile implementation
- 16:48Firebase ML kit overview
- 21:20Building a gesture robot
- 25:15Future trends and takeaways
related talks
39mFrom TensorFlow to ML Kit: power your mobile application with machine learning
29mPushing Machine Learning into your app When TensorFlow and MLKit rule the world
30mMachine Learning in Android Apps with ML Kit
47mMachine Learning on Android Demystified
24mML Kit: Machine learning for mobile developers
38mML Kit: Machine Learning for Mobile with Firebase
42mCreating Smarter Applications
21mGetting the best of on-device ML with ML Kit