Sin descripción

Alexander Freytag b656903ee5 learning is more robust wrt memory issues, less redundant code hace 9 años
data 85c227f3d1 added all relevant files hace 10 años
demos 85c227f3d1 added all relevant files hace 10 años
detect 85c227f3d1 added all relevant files hace 10 años
features 85c227f3d1 added all relevant files hace 10 años
learn b656903ee5 learning is more robust wrt memory issues, less redundant code hace 9 años
misc 85c227f3d1 added all relevant files hace 10 años
LICENSE 85c227f3d1 added all relevant files hace 10 años
README.md bbf4bcef8f markdown support for README file hace 10 años
compileWHOEfficient.m 85c227f3d1 added all relevant files hace 10 años
initWorkspaceWHOGeneric.m 8322dab21d minor bug fix in path setup hace 10 años
test.jpg 85c227f3d1 added all relevant files hace 10 años
train.jpg 85c227f3d1 added all relevant files hace 10 años

README.md

Generic version of WHO for object detection with LDA models

COPYRIGHT

This package contains Matlab source code for object detection with LDA models.

This repo is based on the original version of who from (Bharath Hariharan, Jitendra Malik and Deva Ramanan. Discriminative decorrelation for clustering and classification. In ECCV 2012 ). We significantly altered it, making it more flexible, easier to access, and less focused on hog features in general.

START / SETUP

Compile mex-files via

compileWHOEfficient

Setup workspace

initWorkspaceWHOGeneric

RUN DETECTION DEMO

Run a demo which learns an LDA model for detecting bycicles and using HOG features as underlying representation

demoWhitenedHOG