Learning Part Motion of Articulated Objects Using Spatially Continuous Neural Implicit Representations

Yushi Du*, Ruihai Wu*, Yan Shen, Hao Dong

BMVC 2023

Project

We released the data generation code of Learning Part Motion of Articulated Objects Using Spatially Continuous Neural Implicit Representations here

Installation

1. Create a conda environment and install required packages.

conda env create -f conda_env_gpu.yaml -n PMotion

You can change the pytorch and cuda version by yourself in conda_env_gpu.yaml.

2. Build ConvONets dependents by running python scripts/convonet_setup.py build_ext --inplace.

3. Unzip the data under the repo's root into ./data.

Training

# single GPU
python run.py experiment=Door_emd

# multiple GPUs
python run.py trainer.gpus=4 +trainer.accelerator='ddp' experiment=Door_emd

Testing

# only support single GPU
python run_test.py experiment=Door_emd trainer.resume_from_checkpoint=/path/to/trained/model/

Tuning

# both single and multiple GPUs
python run_tune.py experiment=Door_emd trainer.resume_from_checkpoint=/path/to/trained/model/

Other instructions

You'll also need to follow the instructions here to set up the Earth Mover Distance mentioned in our paper.

bibtex

@InProceedings{du2023learning, author = {Du, Yushi and Wu, Ruihai and Shen, Yan and Dong, Hao}, title = {Learning Part Motion of Articulated Objects Using Spatially Continuous Neural Implicit Representations}, booktitle = {British Machine Vision Conference (BMVC)}, month = {November}, year = {2023} }