CDRL

Requirements

• Install PyTorch (pytorch.org).
• pip install -r requirements.txt
• Download the MEAD dataset from (here).
• Download the pre-trained weights (here) (" backbone.pth ") and place it under "./pretrain/backbone.pth"

Preprocessing

The obtained MEAD dataset is first preprocessed with 'dataloader/align_face.py':

python align_face.py

Get audio feature corresponding to image:

python w2f.py

Training CCRL

cd CCRL 
python train.py

Training CDRL

cd CDRL 
python train.py

The integration of CDRL into NED (Please switch to the NED_CDRL branch.)

First, follow the data preparation and processing workflow of NED.

Supervising the training of the first-stage emotion manipulator

bash run_manipulator.sh

Supervising the training of the second-stage renderer.

bash run_render.sh

Testing metrics on the test set.

cd metrics
bash eval_crossID_driven.sh