Building a live face recognition system
summary
A runnable pipeline shows how to build live face recognition with OpenCV: video capture, release camera or crash, Haar cascades (Viola‑Jones with integral images and AdaBoost, ~160k→~6k features, 32 cascade stages), face cropping + histogram equalization and 50×50 resizing, optional eye‑based alignment, and three recognizers LBPH/Fisher/Eigenfaces (train on saved images, predict with confidence = distance). Notes on practicalities: OpenCV 3.1 vs 3.0 API differences, codec choices (AVI vs MP4/H264), threading for multi‑camera, and a small experiment showing PCA fails on low‑dim tabular data.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 11:27OpenCV image/video basics
- 28:35Face detection with Haar cascades
- 41:10Normalizing face images
- 49:08Eigenfaces and LBPH models
- 59:47Live recognition demo
- 1:04:46Heart disease recognition extension
- 1:11:16Q&A
related talks
40m"Introduction to Face Processing with Computer Vision|"
52mFace Recognition: From Scratch To Hatch
39mBuilding Highly Scalable Facial Recognition Pipelines
01h 11mObject detection using OpenCV
33mDo I Know You? Real time facial recognition with an Apache stack
37mComputer Vision Using Go And OpenCV 3
38m3D Face Tracking and Reconstruction using Modern C++
27mMaking PHP See