Skip to content

Latest commit

 

History

History
47 lines (26 loc) · 1.82 KB

README.md

File metadata and controls

47 lines (26 loc) · 1.82 KB

Recurrent Tracker (RT)

This is our repository of our paper (Fully end-to-end composite recurrent convolution network for deformable facial tracking in the wild).

Requirements :

  1. MTCNN requirements : https://github.com/DuinoDu/mtcnn
  2. Tensorflow GPU : https://www.tensorflow.org/install/install_linux
  3. Other package, can be installed by cloning my environment file on src : env.yml

Landmark Tracker

This main module used to do the tracking for both 2D and 3DA-2D facial landmark. To use :

python testRealtime.py

Also set the is3D in that file to True to have 3D points, False otherwise. Other configuration including filename of video can be set on the configuration.py file on src. Especially to use webcam or the video input.

Example video :

2D Facial Landmark Tracking 3DA-2D Facial landmark Tracking

Real time example :

IMAGE ALT TEXT

Facial Localisation

This module can be used independently to localise facial points from still image. To use :

python facial_localiser.py

Set the is3D in that to be True/False and the filename on the configuration.py.

Some examples :

2D Facial landmark :

Localisation example of 2D landmark

3DA-2D Facial landmark :

Localisation example of 3DA-2D landmark

Citation :

D. Aspandi, O. Martinez, F. Sukno, X. Binefa, Fully end-to-endcomposite recurrent convolution network for deformable facial track-ing in the wild, in: 2019 14th IEEE International Conference onAutomatic Face Gesture Recognition (FG 2019), 2019, pp. 1–8.doi:10.1109/FG.2019.8756630.