Analyzing Video Sequences of Multiple Humans