SPAGHETTI: Editing Implicit Shapes through Part Aware Generation

Installation

git clone https://github.com/amirhertz/spaghetti && cd spaghetti
conda env create -f environment.yml
conda activate spaghetti

Install Pytorch. The installation during development and testing was pytorch==1.9.0 cudatoolkit=11.1

Demo

• Download pre-trained models

python download_weights.py

• Run demo

python demo.py --model_name chairs_large --shape_dir samples

or

python demo.py --model_name airplanes --shape_dir samples

• User controls
  • Select shapes from the collection on bottom.
  • right click to select / deselect parts
  • Click the pencil button will toggle between selection / deselection.
  • Transform selected parts is similar to Blender short-keys. Pressing 'G' / 'R', will start translation / rotation mode. Toggle axis by pressing 'X' / 'Y' / 'Z'. Press 'Esc' to cancel transform.
  • Click the broom to reset.

Adding shapes to demo

• From training data

python shape_inversion.py --model_name  <model_name>  --source training --mesh_path --num_samples <num_samples>

• Random generation

python shape_inversion.py --model_name  <model_name>  --source random --mesh_path --num_samples <num_samples>

• From existing watertight mesh:

python shape_inversion.py --model_name  <model_name>  --mesh_path <mesh_path>

For example, to add the provided sample chair to the exiting chairs in the demo:

python shape_inversion.py --model_name chairs_large --mesh_path ./assets/mesh/example.obj

Training

Coming soon.