Pushing Machine Learning into your app When TensorFlow and MLKit rule the world
summary
Describes building mobile ML pipelines using TensorFlow/TensorFlow Lite and Firebase ML Kit: retraining MobileNet on ~1.7k images for 10 monkey species, converting with TOCO to TFLite (noting version-mismatch issues) and quantizing models to ~10× smaller size. Shows hosting models with Firebase ML Kit and problems with unreliable downloads of ~5 MB models, and presents a face detector that uses ML Kit’s smile probability to trigger jokes while comparing on-device vs cloud accuracy and cost tradeoffs.
AI-written summary · about
chapters (14)
- 0:00ML integration intro
- 1:12Neural networks with TensorFlow
- 4:01Pre-trained models for apps
- 5:24Retrain model with TF Lite
- 8:14Mobile vs Light versions
- 12:29ML Kit for mobile
- 14:55Face detection with ML Kit
- 17:23Host custom model on Firebase
- 18:26ML Kit in Android app
- 21:17Monkey model demo with ML Kit
- 22:40TF Lite and ML Kit limits
- 23:05Why add ML to apps
- 27:16ML app ideas
- 27:38Conclusion
related talks
24mML Kit: Machine learning for mobile developers
39mFrom TensorFlow to ML Kit: power your mobile application with machine learning
29mFrom TensorFlow to ML Kit: power your Android application with ML
30mMachine Learning in Android Apps with ML Kit
36mMachine Learning for Mobile Developers
21mGetting the best of on-device ML with ML Kit
14mDo you want an easy way to add Machine Learning into your app?
09mMachine Learning on the Edge