Headhunter face detector at the oscars


Face detection is a mature problem in computer vision. While diverse high performing face detectors have been proposed in the past, we present two surprising new top performance results.
First, we show that a properly trained vanilla DPM reaches top performance , improving over commercial and research systems.
Second, we show that a detector based on rigid templates - similar in structure to the Viola&Jones detector - can reach similar top performance on this task.
Importantly, we discuss issues with existing evaluation benchmark and propose an improved procedure.


Code and data

Evaluation Toolbox

We propose a new evaluation protocol for face detection (described in section 2 of the paper ). With our evaluation toolbox the figures shown in our paper can be generated and new algorithms can be added for evaluation.
The toolbox contains our updated annotations for the "AFW" and "Pascal Faces" datasets, and the detection bounding boxes of all compared algorithms.

Trained models

We trained several new face detection models. We here provide pre-trained models for our "Baseline" and "HeadHunter" detectors, as well as our "DPM Baseline". Please use DPM version 5 for the DPM baseline and doppia for the other models.

If you use this code or data, please cite

  author = {M. Mathias and R. Benenson and M. Pedersoli and L. {Van Gool}},
  title = {Face detection without bells and whistles},
  booktitle = {ECCV},
  year = {2014}