Evaluation of Methods for Optical 3-D Scanning of Human Pinnas (Detailed Analysis)

This is the detailed analysis of our evaluation as outlined in the paper:
A. Reichinger, Majdak, R. Sablatnig, S. Mairhofer.
Evaluation of Methods for Optical 3-D Scanning of Human Pinnas.
In Proc. International Conference on 3D Vision (3DV 2013), Seattle, Washington, USA, in press.
Pre-print. Final version available at IEEE Explore.

Please note, that the webpages consist of many images and will take some time to load. You can view the "compact" versions without visibility plots, or download the website for faster viewing:
Basic set (41MB), visibility plots (85MB), highres screenshots (165MB), colored PLY meshes (349MB).
Just extract the desired packages into the same directory.

Data sets:

• Subject S1, left ear   (compact)   (Error-coded renderings and statistics for each scan dataset.)
• Subject S1, right ear   (compact)
• Subject S2, left ear   (compact)
• Subject S2, right ear   (compact)
• Subject S3, left ear   (compact)
• Subject S3, right ear   (compact)
• Average   (compact)    (For each method the average over all subjects.)

At the bottom we accumulate the data for all Scanners (averageScanners), all Photogrammetry Methods (averagePS) and All Methods (averageAll).
Most histograms display four subsets stacked on each other (from bottom to top: FO (purple), FI (magenta), BO (brown), BI (green); see clarifications of symbols below).

• Summary Methods    (Boxplots of all datasets in one diagram, combined for each method.)
• Summary Subjects    (Boxplots of all datasets in one diagram, combined for each subject.)

• Plaster ear   (compact)   (Error-coded renderings and statistics for each scan dataset.)

• Ranking.xlsx   (Interactive ranking based on accuracy and completeness. Thresholds and scan-subset can be selected in top left drop-down fields.)


Color coding of signed errors in 3D models:

Clarification of symbols used:

F = front part (see paper)
B = back part (see paper)
O = original parts (not discussed in the paper) which originate from the scanned raw mesh
I = interpolated parts (not discussed in the paper) which were not covered in the scanned raw mesh but completed during the mesh preparation (i.e., hole filling)
F/B and O/I combinations are possible, e.g., OF are only parts which are original AND front.

Ref = Ground-truth reference mesh (GT in paper)
Scan = Scanned mesh after mesh preparation.
Most graphs are given in pairs, the left shows errors from Scan to GT, the right from GT to Scan.

Maintained by Andreas Reichinger, andi <at> vrvis <dot> at. Last updated 24.May 2013.