Sample problem problem001.txt:
problemname VenusStereo1
width 434
height 383
labels 20
labelscale 8
smoothexp 2
smoothmax 7
lambda 50
datacost venus/datacost%03d%c.pgm
datascale 250
smoothx venus/smoothx%c.pgm
smoothy venus/smoothy%c.pgm
smoothscale 1
The venus directory: venus.zip

The parameters used for cost and smoothness images:

match_fn 2   # SSD
match_max 16    # truncated, i.e., cost = (min(diff, 16))^2
match_interval 1 # Birchfield / Tomasi

opt_grad_penalty 1 # no penalty, i.e. smoothx/y images are constant 1