The Multi-Scale Haar Cascade Detection (MSHCD) is an object detection algorithm.
Introduced by Viola-Jones in 2001.
Step 1. Generate raw gray scale image file.
cd ./tools
make
./raw imagefilename
Step 2. Run detection.
cd ./c or ./cpp
make
./mshcd imagefilename cascascadefile
The trained data is gotten from OpenCV project.
And converted into C format that easy to be read in program.
haar_alt.txt
haar_default.txt
File format:
sample-size1 sample-size2
num-of-stages
num-of-trees-at-stage-(i)
/ i j_tree 1 1 x y width height weight
. i j_tree 1 2 x y width height weight
. i j_tree 1 3 x y width height weight //may exist
. i j_tree 1 0 //terminal, not use
. i j_tree 1 j_tree-threshold
. i j_tree 1 left_val
\ i j_tree 1 right_val
......
$i stage_threshold
num-of-trees-at-stage-(i+1)
......
The java implementation in java folder
http://code.google.com/p/jviolajones
The matlab implementation in matlab folder
http://www.mathworks.com/matlabcentral/fileexchange/29437-viola-jones-object-detection
The javascript implementation in javascript folder
https://github.com/foo123/HAAR.js