Detection and Tracking of Multiple, Partially Occluded Humans by Bayesian

kundralaci / JAMResearch

0 stars 0 forks source link

Detection and Tracking of Multiple, Partially Occluded Humans by Bayesian #120

Closed smeyel closed 10 years ago

smeyel commented 10 years ago

DetectionAndTrackingOfMultiplePartiallyOccludedHumansByBayesianCombinationOfEdgeletBasedPartDetectors.pdf Edgelet concept, codebook for edge segments for pedestrian detection.

smeyel commented 10 years ago

Weak classifiers, boosting. Classifiers for separate body parts are learned. Bayesian combination. Detectors separately for 3 possible human orientations. Tracking via data association and mean-shift. Occupancy map: for every pixel the ID of the occupying object. (Useful for visual debugging...)