image - How to combine two features (two minimum distance classifiers) -
hello first post here,
i work on tracking objects through images without prior training. use 2 features, color of region (the ab channels of lab space) , hog. in initial experiments, found using min. distance classifier hog feature alone has advantage of low false positives fp high fn. on other hand, using min. distance classifier color alone increases tp , decreases fn results price of increasing fp.
my question, how combine 2 classifiers? know standard algorithm in unsupervised way.
i tried combine 2 features 1 feature (after normalization) hog dominates results. if weighted combined feature, results worse either of two.
the results reach till (cascade) 2 classifiers, running color first increase possibilities run hog (with threshold little bit higher used hog alone). googled topic don't have enough knowledge classification find standard methods.
thanks
Comments
Post a Comment