Human Motion Transfer with 3D Constraints and Detail Enhancement

Abstract

We propose a new method for realistic human motion transfer using a generative adversarial network (GAN), which generates a motion video of a target character imitating actions of a source character, while maintaining high authenticity of the generated results. We tackle the problem by decoupling and recombining the posture information and appearance information of both the source and target characters. The innovation of our approach lies in the use of the projection of a reconstructed 3D human model as the condition of GAN to better maintain the structural integrity of transfer results in different poses. We further introduce a detail enhancement net to enhance the details of transfer results by exploiting the details in real source frames. Extensive experiments show that our approach yields better results both qualitatively and quantitatively than the state-of-the-art methods.

Dependencies

Jittor
opencv-python
dominate
scipy
tensorflow (for tensorboard)

Dataset

Our approach takes two single-character motion video as input, i.e., the source and target used for motion transfer. The dataeset should be prepared like below

dataset
├──source
|   ├── all_512
|   ├── openpose_img_512_norm
|   └── img_obj_smooth_512_norm
└──target
    ├── all_512
    ├── openpose_img_512_norm
    └── img_obj_smooth_512_norm

Here, the all_512 folder contains the extracted video frames. In our experiment, each frame is cropped to $1024x512$. The openpose_img_512_norm folder is composed of human skeleton images, which are rendered by connecting Openpose[https://github.com/CMU-Perceptual-Computing-Lab/openpose] detecting results. The img_obj_smooth_512_norm folder is consisted of the projection of reconstructed human mesh with MonocularTotalCapture, while the vertex color represents corresponding Laplacian feature. Please refer to the paper for more details.

Training

bash run.sh $source $target

Note that the vairables datadir, codedir, trainname and datasetname can be set in run.sh, since they are always consistent for different characters. The $source and $target denotes the dataset folder name for corresponding characters.

Test

bash test.sh $source $target