About

The OpenHear database is an open database of 3D surface scans of human heads and ears. It was made possible by a research grant from the Oticon Foundation. The 3dMD scanner is placed at the 3D Craniofacial Image Research Laboratory at the University of Copenhagen and we are grateful for their help and support. The acquisition was performed from 2008 to 2010. It is not a stable database and we expect that scans will be added from time to time.

The overall goal of the project is to create a set of scans that can be used for acoustical simulation in for example hearing aid design. For that purpose, we are currently implementing alignment and reconstruction algorithms to create complete head and ear models from our captured data.

Technicalities

The scannings were acquired using a 3dMD cranial scanner in a 2-POD setup. The 3dMD stereo-algorithms are used to create the initial 3D point clouds, while the triangulated surfaces have been created using the Markov Random Field Surface Reconstruction framework. Texture mapping has been done using the texture camera calibration files from the 3dMD calibration. However, in the current scans no texture renormalisation has been done between the two textures. It is therefore possible to see a slight difference in texture between neighbouring triangles.

Currently, the scans are available in VTK format. There is a textured version with no vertex sharing (large number of vertices compared to number of triangles) and a non-textured version with vertex sharing. The non-textured version contains scalar values that are associated to the vertices. The scalar values reflects if the part of the surface is interpolated or based on underlying points (0 - underlying poins, 1 - interpolated).

The scans can be viewed, manipulated, and converted using the free surface manipulation tool Sumatra.

Some of the scans are acquired using a wig cap covering the hair of the person since the scanner is not able to capture hair. Teeths and eyes are also problematic due to their optical properties. Typically the geometry of the eye will be wrong (it turns inward). This is also true for teeth, if they are visible. Facial hair is also problematic as seen on a few of the scannings.

Some textured scans have grey areas. These areas have not been seen by the texture cameras and are therefore likely to be interpolated by the surface reconstruction algorithm.

A short introduction (in Danish) to 3D scanning can be found on Danskernes Akademi.

Terms of use

The data are supplied for scientific purposes on a cite basis. The following reference should be cited:

Rasmus R. Paulsen, Jakob Andreas Bærentzen, Rasmus Larsen, Markov Random Field Surface Reconstruction, IEEE Transactions on Visualization and Computer Graphics, pp. 636-646, July/August, 2010

Contact

Associate Professor Rasmus R. Paulsen
Informatics and Mathematical Modelling
Technical University of Denmark
Richard Petersens Plads
Building 321, office 219
DK-2800 Kgs. Lyngby
Denmark
Phone: +45 4525 3423
Fax: +45 4588 1397
www: http://www.imm.dtu.dk/~rrp
Email: rrp at imm dot dtu dot dk

Data

Frontal textured scans

Male 2 \| zip file	Male 3 \| zip file	Male 4 \| zip file

Male 5 \| zip file	Male 6 \| zip file	Male 7 \| zip file

Male 8a \| zip file	Male 8b \| zip file	Male 8c \| zip file

Male 9 \| zip file	Male 10 \| zip file	Male 11 \| zip file

Female 1 \| zip file	Female 2 \| zip file	Female 3 \| zip file

Female 4 \| zip file	Female 5 \| zip file	Female 6 \| zip file

Female 7 \| zip file

Mannequin 1 \| zip file	Mannequin 2 \| zip file

Boy 1 \| zip file

Girl 2 \| zip file	Girl 3 \| zip file

Fully reconstructed heads

This data is mainly here to show the current status of our reconstruction software. Ear canal scannings are not yet merged with the surface scans.

Female 5 full scan | zip file

Fully reconstructed ears