Abstract: Systems and methods for detecting people in video data streams or image data are provided. The method includes using a plurality of training images for learning spatial distributions associated with a plurality of body parts, detecting a plurality of detections of body parts in an input image, clustering the detections of body parts located within a predetermined distance from one another to create one effective detection for each cluster of detections, and determining a position of each person associated with each effective detection. The detections of body parts can be associated with respective previously learned spatial distributions.

Abstract: Systems and methods for detecting people in video data streams or image data are provided. The method includes using a plurality of training images for learning spatial distributions associated with a plurality of body parts, detecting a plurality of detections of body parts in an input image, clustering the detections of body parts located within a predetermined distance from one another to create one effective detection for each cluster of detections, and determining a position of each person associated with each effective detection. The detections of body parts can be associated with respective previously learned spatial distributions.