Real-time Multi-Facial-Pose-Tracking

Detecting Faces in the ROI,

Focusing on the face close to the center of the video frame,

Detecting 68 Facial landmarks,

Detecting and Tracking facial pose

Setup

To test the code, first make a directory and download the model files in it:

haarcascade_frontalface_default.xml AND lbfmodel.yaml

Make sure to keep the names consitent. Input the directory path to run the code as follows:

Usage

usage: pose_tracker.py [-h] -m MODEL [-i INPUT] [-d DETECTCONF] [-t TRACKCONF] [-s SAVE]

optional arguments:
  -h, --help            show this help message and exit
  -m MODEL, --model MODEL
                        path to directory containing haarcascade_frontalface_default.xml and lbfmodel.yaml
  -i INPUT, --input INPUT
                        path to input video
  -d DETECTCONF, --detectconf DETECTCONF
                        minimum confidence for detection
  -t TRACKCONF, --trackconf TRACKCONF
                        minimum confidence for tracking
  -s SAVE, --save SAVE  Path for saving the output file. If no path is given the file will not be saved

Demo