Aved_Viola - Alex J. Aved 2/14/2005 TOC Overview AdaBoost...

Info iconThis preview shows pages 1–5. Sign up to view the full content.

View Full Document Right Arrow Icon
1 Alex J. Aved 2/14/2005 TOC •Ove rv iew • AdaBoost • Introduction • Detection of Motion
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
2 Objective • Machine Learning approach for pedestrian detection that maximizes detection accuracy and minimizes computation time. • Detection at very low resolution taking advantage of motion and appearance information. Overview • Use machine learning to construct a detector from a large number of training examples. • Work directly with images to detect instances of potential objects. • Use AdaBoost to select a subset of features and construct a cascade of classifiers.
Background image of page 2
3 Overview – Approach • Direct detection of pedestrians – Intensity features – Motion energy features • Feature selection determines optimal balance • Based on Viola/Jones detector (fast) • Overall system is very simple – Large feature set – Training data – Detection requirements – No tracker – No alignment Introduction – Boosting • Consider example of a gambler allowing his agents to make bets on his behalf. • Make a program that predicts accurately the winner of races. • How to combine many rules-of-thumb into an accurate prediction rule. • Boosting is to produce very accurate prediction rule by combining rough and moderately inaccurate rules-of-thumb. AdaBoost – Adaptive boost: Freund & Schapire, AT&T Labs, 1996
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
4 Introduction – Boosting • Booster is provided with a set of labeled training examples • Where yi is the label associated with instance xi ; xi is observable data & yi is the outcome. • On each round t = 1,2 , … ,T , the booster devices a distribution Df over the set of examples, and requests a weak hypothesis ( or rule-of-thumb ) ht with low error with respect to Df. • The distribution of Df specifies the relative importance of each example
Background image of page 4
Image of page 5
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 06/13/2011 for the course CAP 6412 taught by Professor Staff during the Spring '08 term at University of Central Florida.

Page1 / 16

Aved_Viola - Alex J. Aved 2/14/2005 TOC Overview AdaBoost...

This preview shows document pages 1 - 5. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online