Integrated Face Analytics Network
As this code was written using an old version of tensorflow and Keras, a docker is used.
To run the code, build the docker image first by
cd {iFAN_root}/docker
./build_docker.sh
cd {iFAN_root}/
Then start a docker container by
./start_docker.sh
Within the docker container, the {iFAN_root} was mapped to /ifan/.
To run the model which performs face segmentation and landmark localization, go to folder task_seg_pts and run predict.py from there:
cd /ifan/task_seg_pts/
python predict.py
To run the model which performs face segmentation and landmark localization and emotion recognition, go to folder task_seg_pts_emo and run predict.py from there:
cd /ifan/task_seg_pts_emo/
python predict.py
Acknowledgement: Some of the codes to build the Fully Covolutionaly DenseNet are adopted from https://github.com/titu1994/Fully-Connected-DenseNets-Semantic-Segmentation