Through the Looking-Glass Eyes of an Android
summary
Shows concrete results not promises: uses Google Cloud Vision, Microsoft Cognitive Services and IBM Watson for cloud label and face detection (probabilities returned, no face recognition), uses Mobile Vision API for on‑device face tracking and a simple Galileo overlay that works in real time, and uses a Dockerized TensorFlow retrain on Inception to teach a model two classes (Alice vs Stitch) — desktop inference hits ~90% and ~92% for samples, phone inference drops to ~58–78% and retraining on CPU takes hours while production training needs GPUs.
AI-written summary · about
chapters (9)
- 0:00Machine learning on Android
- 3:21Real-world machine learning applications
- 10:59AI terminology and history
- 16:16ML fundamentals with tacos
- 24:35Supervised vs unsupervised learning
- 30:48ML APIs on Android
- 41:14Mobile Vision API
- 44:06TensorFlow and retraining
- 49:19Q&A
related talks
41mMaking Android Apps with Intelligence
40mApplied TensorFlow in Android apps
21mLive Coding: Adding Machine Learning to Android Things in 10 Mins
33mMaking Android Apps with Intelligence
19mOn-device machine learning: TensorFlow on Android
47mMyths and Reality of Android Machine Learning
44mNow smile (also in 3D)! Exploring AR, ML and Camera related APIs on Android
44mHow I Learned to Stop Worrying and Embrace Our Digital Overlords