# This zip file contains the framewise character location data
# for the 50 film clips. Each clip's data is in a separate file
# (NN_character_locations.dat, where NN = clip number 01–50).
#
# In the data files, each line represents one frame (zero-based).
# The location of each character is represented with the x and y
# coordinates (in pixels) of the top-left and bottom-right corners
# of a mask around the character's head. The top-left corner of the
# frame has coordinates (x = 1, y = 1). The dimensions (width and
# height) of the clips for which the data was collected are given
# in the file film_clips.dat. The dimensions can be used to scale
# the data to fit different-sized versions of the same clips.
# The values are comma-separated. In cases where the mask is 
# beyond the frame area, its coordinates are given as "NAN".
# 
# Each line in the data files contains the following values:
#
# 1:  Frame number (zero-based)
# 2:  Shotwise mask index *
# 3:  Clipwise character identifier **
# 4:  Mask top left corner: x coordinate
# 5:  Mask top left corner: y coordinate
# 6:  Mask bottom right corner: x coordinate
# 7:  Mask bottom right corner: y coordinate
# (2–7 repeated for each additional mask in the frame)
# 
# *  Each mask in each shot, irrespective of which character the mask
#    represents, was given a shotwise index. The indices were given in
#    chronological order, starting at 1; i.e. the first-appearing mask
#    in a shot has index 1, the second mask has index 2, etc. Note that
#    the indices only apply within a given shot and bear no relation to
#    the indices in another shot.
# ** Characters appearing in more than one shot were given a unique
#    identifier in chronological order, starting at 1. Incidental
#    characters appearing in only one shot are indicated by the
#    identifier "nil". The dialogue times data is also annotated
#    with the character identifiers.
#
# See the article for more details.