sh env.sh

Download trained weights and required files from here. Unzip it and put it under assets folder.

1. Download the Shape Grading dataset from here.
2. Unzip it and put it under dataset folder.
3. Preprocess the laplacian operater and eignn decomposition of the meshes in the dataset.

   python preprocess/compute_eig.py
   

   This could take a few hours and take up to ~120GB hard drive space to store the results.

python experiments/sota.py 

cd train_weights
python main/train.py --config_file configs/debug.yaml