We just updated our face detection paper project page to include the results from the recently introduced Multi-Attribute Labelled Faces (MALF) dataset.
Our ECCV2014 methods reach top performance on all major face detection benchmarks: AFW, Pascal Faces, FDDB, and MALF.
We also added the link to presentation video.
The main point of the methods proposed in ECCV2014 is that these are baselines, inspired by 5~10 years old methods (DPM and Viola&Jones). Looking forward to see how much face detection will further improve in the next iterations.