README.md 718 B

opencv-hog

  • extracts feature vector from image (see feature_vector function)
    • "simple" hog
    • hog_cv
    • color histogram
  • support vector machines classification -- training set (first 80% of data) generates model -- testing set (last 20% of data) evaluates model
  • list of true/false positive/negative of testing set is printed in html to stdout

The input is defined as set of pictures with same resolution (e.g. 48x48) and class name

annotation file looks like this:

/var/www/html/fire/sun-dataset/SUNOUT/a/art_school/sun_aprywzqxjcezkbch.jpg,non-fire
/var/www/html/fire/labelme_samples-80/000093_65.jpg,fire

Sample Output