ItMRnet

An iterative solution for improving the generalization ability of unsupervised skeleton motion retargeting

>>>>>>>>>Downloading and Preprocessing Data<<<<<<<<<
All data we use are come from the website of mixamo(https: //www.mixamo.com/ ). You need to create an account to get the animation files. There is a way to quickly start downloading files.
> *https://github.com/gnuton/mixamo_anims_downloader*
Once the fbx animation files have been downloaded, run *fbx2bvh* in Blender(note: press *"Shift + F11"* to jump to the script screen). Then, process the bvh files into npy file by *bvh2npy.py*.

>>>>>>>>>Training Phase<<<<<<<<<
Training ItMRNet by:
> *./src/ItMRnet_Train.py*
The model will be saved at "./models". We also provide pre-trained models to get the help started quickly(models are saved at "./models").

>>>>>>>>>Execution Phase<<<<<<<<<
Executing ItMRnet by:
>*./src/ItMRnet_Test.py*
The details of the four groups of experiments are shown at *https://drive.google.com/file/d/1T0o6kFho96CniyjX7LXV_xgaeHlDJqbV/view?usp=sharing*.
The experimental data are shown at "./datasets/test/"

>>>>>>>>>Evaluation<<<<<<<<<
> Run* ./result/evaluate.py*

>>>>>>>>Rendering model to generate videos<<<<<<<<<
Download blender files with character skins -->https://drive.google.com/file/d/1m_n1RzdmOv8o1KQJpH3YDC0Drb7KcH8Z/view?usp=sharing
>Run *./result/make_videos.py* to generate videos.
The visualization results of ItMRNet are all placed in the video at https://drive.google.com/file/d/15Jv_qdcdsS8Rn26lIZc-o5T8YaKgi17R/view?usp=sharing.

>>>>>>>>>IK processing<<<<<<<<<
Please refer to this work：*https://github.com/DeepMotionEditing/deep-motion-editing*