McGillFaces Database

 


Introduction

Welcome to McGill Real-World (Unconstrained) Face Video Database, a database of video sequences collected for studying the problem of unconstrained face classification from videos.

This database contains 18000 video frames of 640x480 resolution from 60 video sequences, each of which recorded from a different subject (31 female and 29 male). Each video was collected in a different environment ( indoor or outdoor) resulting arbitrary illumination conditions and background clutter. Furthermore, the subjects were completely free in their movements, leading to arbitrary face scales, arbitrary facial expressions, head pose (in yaw, pitch and roll), motion blur, and local or global occlusions.

Explore

See the video below showing a collection of short clips from some of the subjects.


Data Format


NEWS!!! The face masks are now available to download. Please email me if you have already downloaded the database and would like to download the masks.

For each video (subject), we provide 
  • Original video frames
  • Detected, aligned, scaled and masked face images after applying the preprocessing step introduced in [1]
  • Gender ground truth
  • Facial hair ground truth
  • Continuous and probabilistic head pose (yaw) ground truth obtained via the manual labelling method in [1]

The chosen naming convention for the video frame is of the form aa_b_ddd.jpg  where

aa        = subject ID

b           = gender label, where 0=male and 1=female

ddd      = frame ID, where ranges from 001 to 300.


Check the README.txt for the details of the files provided in our database.

Download

This database is only for non-commercial (academic) use. 

Note that currently a subset of the database is available due to the ongoing consent process. As more subject are released, subscribers will receive update emails from me. Thank you for your patience.

References

If you use the database and/or use the pose labels, please make sure to cite:

M. Demirkus, D. Precup, J. Clark, T. Arbel, "Hierarchical Temporal Graphical Model for Head Pose Estimation and Subsequent Attribute Classification in Real-World Videos"Computer Vision and Image Understanding (CVIU), Special Issue on Generative Models in Computer Vision, March 2015. 

M. 
Demirkus, J. J. Clark and T. Arbel, "Robust Semi-automatic Head Pose Labeling for Real-World Face Video Sequences"Multimedia Tools and ApplicationsJanuary 2013. 

 

If you do the pose/gender/facial hair analysis on this database, please cite the corresponding paper:

M. Demirkus, D. Precup, J. Clark, T. Arbel, "Hierarchical Spatio-Temporal Probabilistic Graphical Model with Multiple Feature Fusion for Estimating Binary Facial Attribute Classes in Real-World Face Videos"IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), Sept. 2015. 

M. Demirkus, D. Precup, J. Clark, T. Arbel, "Hierarchical Temporal Graphical Model for Head Pose Estimation and Subsequent Attribute Classification in Real-World Videos"Computer Vision and Image Understanding (CVIU), Special Issue on Generative Models in Computer Vision, March 2015. 

M. Demirkus, D. Precup, J. Clark, T. Arbel, "Probabilistic Temporal Head Pose Estimation Using a Hierarchical Graphical Model", European Conference on Computer Vision (ECCV), 2014.
 


Contact
Questions and comments can be sent to:
Meltem Demirkus