• 设为首页
  • 点击收藏
  • 手机版
    手机扫一扫访问
    迪恩网络手机版
  • 关注官方公众号
    微信扫一扫关注
    迪恩网络公众号

cheeyi/matlab-viola-jones: A slightly modified version of Viola-Jones face detec ...

原作者: [db:作者] 来自: 网络 收藏 邀请

开源软件名称(OpenSource Name):

cheeyi/matlab-viola-jones

开源软件地址(OpenSource Url):

https://github.com/cheeyi/matlab-viola-jones

开源编程语言(OpenSource Language):

MATLAB 100.0%

开源软件介绍(OpenSource Introduction):

Viola-Jones Face Detection for Matlab

A CSCi 5561 Spring 2015 Semester Project

Authors: Chee Yi Ong, Stephen Peyton

Introduction

This is a slightly modified Viola-Jones face detection algorithm built using Matlab. Here's a quick rundown of the code flow:

  • Preprocessing: variance normalization, gamma correction for ‘hard’ (under/over-exposed) images
  • Train weak classifiers from Haar-like features
  • Boost weak classifiers using Adaboost
  • Face detection using a cascade structure

Assumptions

  1. Frontal-facing images ONLY.
  2. Background is not cluttered. Solid-colored background works the best.
  3. Tilting of the head is at a minimum.
  4. Image size is approximately 300x400 or similar. Individual features are a minimum of 19x19, because that is the smallest size of a single Haar feature or classifier.
  5. One face-of-interest per image.

Instructions

This folder contains two subfolders: trainHaar and detectFaces. trainHaar consists of the training algorithm which trains classifiers using Haar-like features, while detectFaces uses the trained classifiers to detect faces.

The main functions for both parts of the face detection routine are named identically to the folder containing the code, i.e., trainHaar.m for the training part, and detectFaces.m for the detection part.

  1. Training: simply start the training by running the script trainHaar on the command line. Note that this takes approximately 21 hours on a 2.6GHz quad-core i7.
  2. Detection: detectFaces('image.jpg') or detectFaces('someDirectory/image.jpg').

Opportunities for improvements:

  • Train algorithm with a larger set of images
  • Better thresholding with more Adaboost training rounds
  • Better cascade structuring with fewer, stronger classifiers: real-time detection possible

Acknowledgements

  • University of Minnesota, Twin Cities
  • Viola, Paul, and Michael J. Jones. “Robust real-time face detection.” International journal of computer vision 57.2 (2004): 137-154.
  • Freund, Yoav and Schapire, Robert E.. “A decision-theoretic generalization of on-line learning and an application to boosting.” Second European Conference, EuroCOLT ’95, pages 23–37, Springer-Verlag, 1995.
  • Anila, S. and Devarajan N.. “Preprocessing Technique for Face Recognition Applications under Varying Illumination Conditions.” Global Journal of Computer Science and Technology 12.11-F (2012).
  • MIT Center for Biological and Computational Learning. “CBCL Face Database 1”. N. p., 2015. Web. Accessed 16 April 2015. http://cbcl.mit.edu/software-datasets/FaceData2.html
  • “AT&T Face Dataset”, http://www.cl.cam.ac.uk/research/dtg/attarchive/facedatabase.html



鲜花

握手

雷人

路过

鸡蛋
该文章已有0人参与评论

请发表评论

全部评论

专题导读
热门推荐
阅读排行榜

扫描微信二维码

查看手机版网站

随时了解更新最新资讯

139-2527-9053

在线客服(服务时间 9:00~18:00)

在线QQ客服
地址:深圳市南山区西丽大学城创智工业园
电邮:jeky_zhao#qq.com
移动电话:139-2527-9053

Powered by 互联科技 X3.4© 2001-2213 极客世界.|Sitemap