close

Вход

Забыли?

вход по аккаунту

код для вставкиСкачать
Week 5
Emily Hand
UNR
AdaBoost

For our previous detector, we used SVM.


Color Histogram
We decided to try AdaBoost

Mean Blocks
Mean Blocks

8x8 pixel blocks inside of
template

Take mean R, G, and B
values for each block and
make a feature vector
Bad Results
Continue SVM
Continue SVM
Continue SVM

New search area around detector

Threshold for learning

This is only with color histograms
–
We will use other features as well
TLD Tracker
TLD Tracker


Room for improvement.
–
Haar-like features
–
Adaboost Detector
Problem with their detector

Strong detection far from the tracker may cause
problems
Next Week

Train with new features
–
Haar Features
–
LBP
–
HOG

Optical Flow...

Extract detector from TLD code
1/--страниц
Пожаловаться на содержимое документа