Skip to content
/ SALS Public

[ICLR 2025] Shape as Line Segments: Accurate and Flexible Implicit Surface Representation

Notifications You must be signed in to change notification settings

rsy6318/SALS

Repository files navigation

SALS (ICLR 2025)

Shape as Line Segments: Accurate and Flexible Implicit Surface Representation [Paper]

Preparation

Install ngpmesh to do the intersection detection.

cd third_party/ngpmesh
pip install .
cd ../..

Generate E-DC data for surface extraction. It will take a while, please be patient. (Welcome to discuss more efficient solutions with us!!!)

python neural_shape_representation/diff_emc.py

Neural Shape Representation

Utilize an MLP to model a shape, and output the extracted surface.

cd neural_shape_representation
python overfit_main.py --filename <input_mesh> --ckpt_path <save_path>

Point Cloud Surface Reconstruction

Prepare dataset.

python preprocess_data.py

Train the network.

python train.py --data_path <dataset path>

Evaluate the network

python eval.py --config_path <path to config file>

About

[ICLR 2025] Shape as Line Segments: Accurate and Flexible Implicit Surface Representation

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages