XM2VTS 68pt Markup

Introduction

The XM2VTS frontal data set contains 2360 mug shots of 295 individuals collected over 4 sessions. More information on the data set can be found at:-

http://www.ee.surrey.ac.uk/Research/VSSP/xm2vtsdb/

Example images are shown below:-

egs1	egs2	egs3	egs4

Manual Landmarking

The XM2VTS data set is supplied with manually located eye points for all 2360 images. These points files can be obtained from the University of Surrey web site, here.

However to enable more detailed testing and model building the XM2VTS markup has been expanded to landmarking 68 facial features on each face. The 68 points chosen are consistent across all images. The landmark scheme is shown below:-

Click on the image or here to see a larger version of this image

It is intended that these marked up point will be used for a variety of purposes:-

• Facial Feature Finding - The markup provides ground truth to test automatic face and facial feature finding software.

• Face Model Building - Sophisticated object models, such as the Active Appearance Model approach require manually labelled data, with consistent corresponding points as training data.

• Face Recognition/Verification - The marked up points could be used to test face recognition or verification performance, when the exact location of the face and features are known (i.e. removing the registration error)

Points Data

Currently the full 68 point markup is available for sessions 1-4.

xm2vts_session1.zip

xm2vts_session2.zip

xm2vts_session3.zip

xm2vts_session4.zip

The filenames of the XM2VTS images have the following format:-

XXX_Y_Z.ppm

Here:-

• XXX= A number from 000-371 indicating the identity of the subject
• Y= The session number (1,2,3 or 4)
• Z= 1 or 2, two shots were taken at each session, Z indicates which one

The image files and point files have corresponding names. e.g. "000_1_1.ppm" has points in file "000_1_1.pts"

The points file format is as follows:-

version: 1          (This points file version no. can be ignored)
image_size_x: 720   (The image width in pixels)
image_size_y: 576   (The image height in pixels)
n_points: 68        (The number of labelled points on the image)
{
xxxx  yyyy
.....
}

For each point xxxx is the x co-ord starting from the top-left corner and yyyy is the y co-ord similarly starting from the top-left corner of the image.

All points files contain 68 points with each point representing a specific point on the face (see diagram above).

Acknowledgement