Neural Implicit Shape Editing using Boundary Sensitivity

Arturs Berzins, Moritz Ibing, Leif Kobbelt

arXiv | Presentation & Poster | OpenReview

GUI Demo for Semantic Editing

The GUI is based on the DualSDF demo. The method works with different models, given a proper interface to the model to evaluate it.

Three pretrained models are available under interfaces/DualSDF/pretrained/ based on the DualSDF generator: airplanes, cars, and chairs.

# Host the chair demo on the default port 1234
python demo.py interfaces/DualSDF --interface_args \
"./config/dualsdf_chairs_demo.yaml --pretrained ./pretrained/dualsdf_chairs_demo/epoch_2799.pth"

Then visit localhost:1234. Type python demo.py --help to see and adjust the launch arguments.

demo.mp4

TODO

• Add pre-trained IM-Net.
• Details on how to add your own model.
• Host the large pretrained model files not on GitHub

Licence

MIT License