3D Object Samples
Stereo
These are snapshots of a cup. Each sample is approximately 20K points at 1m, more when closer. Texture was added to the cup to get rid of dropouts. Check out the inside views! The attachments can be viewed in a web browser using a VRML plugin. Cortina works well with MS Windows, and FreeWRL is good for Linux.
9 cm baseline, monochrome, 1m distance
Cup 1st side view download
Cup 2nd side view download
Cup 3rd side view download
Cup inside view download
Cup top view download
6 cm baseline, 5.6 mm focal length, color, various distances
Cup 35cm download
Cup 35cm download
Cup 35cm download
Cup 50cm download
Cup 50cm download
Cup 100cm download
Cup 100cm download
Cup 150cm download
6 cm baseline, 3.0mm focal length, monochrome, various distances
Cup 30cm download
Cup 100cm download
Cup+Table 50cm download
Cup+Table 50cm visual download
Stereo with Random Pattern
6 cm baseline, 5.6 mm focal length, color, various distances
This scene shows the potential of projected random patterns. The pattern (see visual below) is generated using a uniform probability of b/w at each square; better patterns might emphasize vertical texture. The scene is difficult, with very dark objects (the chair) and very reflective specular objects (the cup).
Chair+cup 100cm download
Chair+cup 100cm visual download
Tilting Hokuyo
Office download
Office with planar segmentation download
Office with Intensity download
Cup download
Cup on its side download
VRML Point cloud with intensity data Raw points download
Tilting Scan Video (MP4) download
Veiling Effect in Hokuyo UTMs
Courtesy John Spletzer at Lehigh University. They have a UTM-30LX, not a UTM-X001S, but the effect appears to be the same.
UTM-30LX.jpg - There is a foreground object at 0.9m and a background object at 2.0m. The interpolated points are incorrect.
Checkerboard Pairs
See Scanner to camera calibration for more info on calibration
- Hokuyo and camera data pairs and calibration files
- Hokuyo IR intensity image (spherical projection)
- Camera image
SR4K Images
The Swiss Ranger 4K model is a flash ladar with resolution of 176 x 144 pixels. Here are some depth and intensity images from the device, with fixed pattern noise correction and 7x7 adaptive neighborhood filter. Thanks to Radu Rusu of TUM.
Tabletop view @ 1m depth download
Tabletop view @ 1m intensity download
Depth Images
See /wg/stor1/data/08-07-08 Stereo For Borg Classifier/
xx-stereo.txt and xx-intensity.txt contain the output of SVS's "Store 3D Point Array" for with the textured lighting on and off, respectively. Exposure was manually adjusted to be right for each image. Format is:
X Y Z r g b [imgx imgy] disparity
xx-combined.txt contains data in nearly the same format but missing the disparity value at the end of each line; it has X Y Z from xx-stereo.txt and r g b from xx-intensity.txt.
xx-depth.txt is a 640x480 ascii matrix of depth (Z) values, as this is the format that is used by the borg classifier.