The source code (pytorch) for our project: Image-to-Video Generation via 3D Facial Dynamics.
We use the IJBC video datasets for the training of our generation model. All the images are cropped by the 2DAL algorithm. The face video folders with poor quality are removed firstly, 2447 video folders are selected for the training & testing. 20 video folders are randomly selected for testing, others are used for training. The training and testing subsets can be downloaded at URL: https://pan.baidu.com/s/1MgNu593209IrQQzKq5XPXw Password: cfu3
- Download the files at URL: https://pan.baidu.com/s/1CVRdw5JaMrtKzA8BXmzZuA Password: zlcp
- Extract the files under the root folder.
- Run the main script "main_paraltrain_withMask_300vw_idloss" for training, all the implementation details are contained in this script.