Neural Wavelet-domain Diffusion for 3D Shape Generation, Inversion, and Manipulation [ACM TOG 2023]

This is the code for "Neural Wavelet-domain Diffusion for 3D Shape Generation, Inversion, and Manipulation"

Authors: Jingyu Hu*, Ka-Hei Hui, Zhengzhe Liu, Ruihui Li, Chi-Wing Fu (* joint first authors)

Environment

For environment setup, we follow Wavelet-Generation.

Dataset

For the dataset, we use the train/test split provided by IM-NET for training and evaluation. We provided the pre-computed data in this link: https://drive.google.com/file/d/1hGTDRLqf8GhCy5wnCGWIOxrIZ1bia9ZE/view?usp=sharing

Training

To be released soon

Inference

We provided our pre-trained models for shape inversion with the following link: https://drive.google.com/file/d/1wzUe35XclNQOdusdtwy3TP65jgIhg9Os/view?usp=sharing

You can download the pre-trained models and put it in the "./testing_folder"

You can run the inference by running:

python eval_con_diffusion.py