Card detection using YOLO on Android
summary
Builds a playing‑card detector using Tiny YOLO v2 trained in Darknet, converts weights to TensorFlow then to .tflite, and runs on Android via the TF Lite interpreter. Uses about 12–15 images per card for 52 classes, trains on an NVIDIA GPU for usable results, and annotates images manually (augmentation recommended). Reports ~200 ms inference for a small class set, slower with all 52 cards, and roughly 90% accuracy on plain backgrounds dropping to ~60% in varied scenes.
AI-written summary · about
chapters (7)
- 0:00Start
- 3:31What's YOLO
- 10:18Output
- 10:59Non-max suppression
- 13:35Process
- 15:25Generate dataset
- 18:44Train YOLO
related talks
40mWho’s afraid of Machine Learning? First steps with TensorFlow for Android.
47mDeep Learning: a practical introduction from an Android dev perspective
19mOn-device machine learning: TensorFlow on Android
15mSimplified Machine Learning on mobile, browsers, and edge devices with on-device ML
42mCreating Smarter Applications
32mApplying Deep Learning for Object Detection from a Telecom Perspective
39mAI/ML on mobile/android with TensorFlow
39mMachine Learning on Android Demystified