"MPI Dancer"
A test sequence for multi-view reconstruction
and codec research

Introduction
Download
Additional notes and news
Contact information

Introduction

This is the web-home for GrOVis' FVV texture codecs. The FVV texture codecs utilize new combinations of known compression techniques to yield more compression out of 4D (time, camera angle, image dimensions) texture data that is commonly used in model-based FVV.

DOWNLOAD: PDF of PowerPoint slides used in a recent lecture on this topic.

The data is meant to be used as an example for what the codecs are able to achieve.

The set consists of frame sequences rendered from 8 different camera views located in a circle around the scene. The recorded scene contains a humanoid figure, animated by motion-captured data. A shilhouette set can be extracted with an RGB != (0,0,0) test. The sequences are 465 frames long for the texdiff codec and 40 frames long for the 4dspiht codec. Each frame is saved as a 512x512 image. Some snapshots from the sequences can be seen in Figure 1.

Figure 1. Three frames, same timestep, from different camera views of the set.
Underlying geometry. Rendered view with two smaller original recordings.

Download

We currently only provide the files separately, from HTML pages, FTP and tar-files will follow.
A good solution for the download is to cut & paste the directory name into wget, e.g.

wget -l 1 -r http://www.mpi-sb.mpg.de/~gziegler/texfold_data/soutput_070/

Please be advised that the texdiff directories are HUGE!

The Data set (single files)

FileDescriptionSize
output_originalUncompressed, frame 65-462.125 MB
output_005texdiff, 0.05 bpp, frame 65-462.125 MB
output_025texdiff, 0.25 bpp, frame 65-462.125 MB
output_070texdiff, 0.70 bpp, frame 65-462.125 MB
soutput_0054DSPIHT, 0.05 bpp, frame 100-139.7 MB
soutput_0254DSPIHT, 0.25 bpp, frame 100-139.7 MB
soutput_0704DSPIHT, 0.70 bpp, frame 100-139.7 MB

The Data set files (.tar.gz format, directory names)

FileDescriptionSize
*.tar.gzTar files.125 MB

The Calibration set (.tar.gz format)

FileDescriptionSize
dancer_calib_data.tar.gz.dat files and README in .tar.gz2 kB

Notes and news

Contact Information

If you have any questions or comments you are welcome to send them to Gernot Ziegler, gziegler@mpi-sb.mpg.de.

www.grovis.de
Graphics-Optics-Vision
Max-Planck-Institut für Informatik
Stuhlsatzenhausweg 85
D-66 123 Saarbrücken
Germany

Several members of the Graphics-Optics-Vision group have been involved in the construction of this data, including Christian Theobalt, Naveed Ahmed, and Marcus Magnor.