Orthogonal projection images for 3D face detection

This is an official implementation for Orthogonal projection images for 3D face detection implemented in C++ and based on OpenCV.

Introduction

This is a real-time 3D face detector based on boosted cascade classifiers that uses a scale-invariant image representation to improve both efficiency and efficacy of the detection process, named orthogonal projection images. In this representation, images are not scanned at multiple scales in order to detect faces with different distances in relation to the camera. It achieves a high degree of pose invariance by detecting frontal faces not only in the camera viewpoint but also in rotated views of the scene.

Requirements

• g++ compiler with support to C++11;
• OpenCV, from the repository. We tested it up to version 3.2.0;
• [Optional] librealsense. Only necessary to run the demonstrations using the Intel® RealSense™ acquisition devices.

Installing the demonstrations

We only tested it with Ubuntu 16.04, but any other Linux distribution should be fine. To compile the point cloud demos:

• mkdir build && make

Then, to compile the Intel® RealSense™ demos:

• make realsense

Or the .abs compatible (FRGC's chosen format) ones:

• make frgc

Usage

The point cloud demonstrations assume that the point_cloud_demo.cpp (or point_cloud_frontal_demo.cpp) will be modified so that a point cloud in format std::vector<cv::Point3d> will be read in points prior to calling the methods. The generated executable are called demo.out and frontal_demo.out. Coordinates must be in milimeters throughout the code.

FRGC

FRGC's compatible demos are run by:

• ./frgc_demo.out $path_to_abs_file
• ./frgc_frontal_demo.out $path_to_abs_file

The detected face will be shown and marked in a blue rectangle. For example, you can use the abs file in the samples folder:

./frgc_demo.out samples/face.abs

Intel® RealSense™

Intel® RealSense™ demos require only running the executable ./realsense_demo.out or ./realsense_frontal_demo.out. The detected face will be marked by a blue rectangle in 30FPS realtime.

Citing

If you find the code in this repository useful in your research, please consider citing:

@article{PAMPLONASEGUNDO201472,
  title = "Orthogonal projection images for 3D face detection",
  journal = "Pattern Recognition Letters",
  year = "2014",
  author = "M. Pamplona Segundo and L. Silva and O.R.P. Bellon and S. Sarkar",
}

Documentation and Troubleshooting

While the code is still not fully documented, there are comments in each method relevant to the implementation detailing its purpose, parameters and return value. Questions and problems can be reported as issues in this repository. We would be glad to answer them.