Deep Learning: a practical introduction from an Android dev perspective
summary
An Android dev walks through installing Anaconda and TensorFlow 2.0 alpha, wrestling with CUDA and PyCharm, then ports a pretrained DeepLab v3_257.tflite model to TensorFlow Lite in Kotlin, mapping 257×257×3 float inputs to 21×257×257 probability outputs and running it via CameraX (Pixel 3 is slow; Pixel 4 much faster). A second project trains a classifier from two million chat messages using MD5 hashing, 50‑word padding and an embedding layer to predict one of five emotes, showing data choices matter far more than model tweaks.
AI-written summary · about
chapters (8)
- 0:00Introduction and motivation
- 4:31AI vs machine learning vs deep learning
- 10:41Setting up the ML environment
- 13:30Building an image segmentation app
- 23:06From using models to building one
- 33:02Text encoding for deep learning
- 40:23Data quality and model training
- 43:35Key takeaways for Android devs
related talks
42mCreating Smarter Applications
01h 04mWhat you always wanted to know about Deep Learning, but were afraid to ask
40mWho’s afraid of Machine Learning? First steps with TensorFlow for Android.
41mWarm Up Your Smartphone's Neurons with Deep Learning
40mApplied TensorFlow in Android apps
39mAI/ML on mobile/android with TensorFlow
47mMachine Learning on Android Demystified
49mDeep Dive into Deep Learning with KotlinDL