C.S. Pilz, S. Zaunseder, J. Krajewski, V. Blazek,
Local Group Invariance for Heart Rate Estimation from Face Videos in the Wild,
The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, pp.1254-1262, Salt Lake City, 2018
We recorded four different sessions. The first session consists of a resting scenario where no head or facial motion is performed and the illumination is more or less static. In the second session the users are asked to perform head as well as facial motions but the illumination remains static. The third session is performed during an exercise on a bicycle ergometer in a gym where no further instructions are given to the user. The fourth sessions is recorded during an urban conversation including head and facial motions as well as natural varying illumination conditions. The images above show some example images taken from the recordings during the different sessions. Every session is recorded over an one minute time span, except the ergometer session which is recorded over a 5 minute period. In total 25 users were ask to participate resulting in an amount of 100 video recordings with approximately 200 min total duration; more than three hours. The data collection consists of 20 male and 5 female in the range of 25-42 years. The majority ethnicity is Caucasian. The camera device is selected as Logitech HD C270 webcam and as reference ground truth measurements we synchronized a common finger pulseoximeter, a CMS50E PPG device, over its serial port communication protocol. The average frame rate of the camera is set to 25 FPS and for the pulseoximeter 60 FS. The camera video stream is captured uncompressed with auto-exposure and stored into an AVI container. For every captured image the time stamp is stored too. The pulseoximeter signal is stored together with the device’s pre-computed heart rate information.
NOTE:
Due to ongoing bandwidth hustle,
there are no sessions for dowload available currently.
It seems the company isn't willing to provide storage and
bandwidth for hosting the LGI videos anymore.
I'm sorry that such relatively simple and sensible things are torpedoed by various parties.
Since I've rescued the first six sessions I might be able to provide these for the community
in case someone has enough cloud storage where I could upload the zip archives.
In the mean time i'm going to setup a small server at home for hosting the sessions.
Or try to contact the company CanControls in Aachen/Germany.
info@cancontrols.com
The data can be downloaded using the following links:
(i'll update the missing links as soon as i got some more webspace!)
- ID7 t.b.c.
- ID8 t.b.c.
- ID9 t.b.c.
- ID10 t.b.c.
- ID11 t.b.c.
- ID12 t.b.c.
- ID13 t.b.c.
- ID14 t.b.c.
- ID15 t.b.c.
- ID16 t.b.c.
- ID17 t.b.c.
- ID18 t.b.c.
- ID19 t.b.c.
- ID20 t.b.c.
- ID21 t.b.c.
- ID22 t.b.c.
- ID23 t.b.c.
- ID24 t.b.c.
- ID25 t.b.c.