2001 Stereo datasets with ground truth

These datasets of piecewise planar scenes were created by Daniel Scharstein, Padma Ugbabe, and Rick Szeliski. Each set contains 9 images (im0.ppm - im8.ppm) and ground-truth disparity maps for images 2 and 6 (disp2.pgm and disp6.pgm). Each ground-truth disparity map is scaled by a factor of 8. For example, a value of 100 in disp2.pgm means that the corresponding pixel in im6.ppm is 12.5 pixels to the left.

Click on the images to see an animated gif of each sequence. Click on the sequence name to download the images.
Sawtooth Venus Bull Poster Barn 1 Barn 2

Only the sawtooth and venus images are used in our study, together with the following two sequences: Tsukuba (courtesy of the University of Tsukuba) and Map (courtesy of R. Szeliski and R. Zabih)
Map Tsukuba

Animated gifs of the reprojected sequences based on ground truth:

Inverse warping (all images are warped to the reference frame viewpoint):
Half size: Tsukuba(140K) Sawtooth(400K) Venus(370K) Map(22K)
Full size: Tsukuba(440K) Sawtooth(1.3M) Venus(1.2M) Map(65K)

Forward warping (the reference frame is warped to all other viewpoints):
Half size: Tsukuba(140) Sawtooth(400) Venus(370K) Map(22K)
Full size: Tsukuba(440K) Sawtooth(1.3M) Venus(1.2M) Map(69K)

Additional test images (mostly without pixel-accurate disparity maps) are available from the Computer Vision Homepage.

Last modified: Fri Feb 11 02:06:35 EST 2011 by Daniel Scharstein