Thin plate spline motion model (TPSMM) converted to ONNX
Update 2024-05-01
Faster inference code when running on GPU
Bug fix onnx conversion. removed bg_predictor.
driving - old version - new version:
version2.mp4
.
run inference:
python demo.py --source source.png --driving driving.mp4 --output output.mp4
optional parameters: --mode standard (default relative) and --cpu to run on cpu (default cuda)
I converted the models using modified code from https://github.com/TalkUHulk/Image-Animation-Turbo-Boost
Original TPSMM repo : https://github.com/yoyo-nb/Thin-Plate-Spline-Motion-Model