peterzs / idf

implicit displacement field

Geometry-Consistent Neural Shape Representation with Implicit Displacement Fields

[project page][paper][cite]

• Geometry-Consistent Neural Shape Representation with Implicit Displacement Fields
  • overview
  • demos
    • preparations
    • surface reconstruction
    • detail transfer
  • bibtex

overview

demos

cuda 11.1 and pytorch 3.8

preparations

git clone https://github.com/yifita/idf.git
cd idf

# conda environment and dependencies
# update conda
conda update -n base -c defaults conda
# install requirements
conda env create --name idf -f environment.yml
conda activate idf

# download data. This will download 8 mesh and point clouds to data/benchmark_shapes
sh data/get_data.sh

surface reconstruction

# surface reconstruction from point cloud
# replace {asian_dragon} with another model name inside the benchmark_shape folder
python net/classes/runner.py net/experiments/displacement_benchmark/ablation/ablation_phased_scaledTanh_yes_act_yes_baseLoss_yes.json --name asian_dragon

detail transfer

This example uses provided base shapes

sh data/get_dt_shapes.sh

# evaluation of the pretrained examples. This will save the results in 'runs/shorts_residual_filmsiren'
python net/classes/runner.py net/experiments/transfer/shorts_2phase.json

Or you could also train these examples yourselves:

sh data/get_dt_shapes.sh

# this will train the base shapes for the source and target shapes, then train the transferable idf
python net/classes/executor.py net/experiments/transfer/exec.json

bibtex

@misc{yifan2021geometryconsistent,
      title={Geometry-Consistent Neural Shape Representation with Implicit Displacement Fields},
      author={Wang Yifan and Lukas Rahmann and Olga Sorkine-Hornung},
      year={2021},
      eprint={2106.05187},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}