Teaching the iPhone to See
summary
Prototype uses OpenCV Cascade classifier on-device to read game UIs, with a practical pipeline: color-filter the red channel, capture tens of thousands of examples, train offline and ship an XML model. A three-week effort yields a working iPhone app that detects Counter-Strike radar blips and League of Legends minimap enemies, requiring roughly 20,000 positive samples and similar negatives, and running fast enough to drive haptic alerts in real time.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 0:34Machine Learning
- 0:53Perspective
- 3:54Machine Learning Applications
- 5:08Why Im Excited
- 8:56Sunbreaker Titan
- 9:25Hammer
- 9:42CounterStrike
- 13:02Proof of Concept
- 13:37Getting Started
- 15:39OpenSeaorg
- 18:43Cascade classifier
- 20:10Demo
- 21:32Code Walkthrough
- 24:22DetectMultiscale
- 25:43Radar Car
- 26:02Data
- 26:49Results
- 27:37Capturing Data
- 28:13Folders
- 28:34Filtering
- 30:59Edge Detection
- 32:01Mission accomplished
- 32:33Image size
- 35:14red circles
- 36:59examples
- 37:20excitement
- 39:57QA
related talks
33mComputer Vision, Not Just for Breakfast Anymore
38mDeep Learning for detection on a phone
40mDeep Learning on iOS
40mNavigating the Machine Learning Landscape for Apple Devices
20mTaming the Machine (feat. Nicolas Cage)
55mCore ML: Vision
34mThe Potential of IMAGE Processing in Swift for Developers
10mMachine Learning And React Native