Skip to content

Simple MLP for representing the SDF of a single shape

Notifications You must be signed in to change notification settings

nihalsid/shape_sdf

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SDF representation with an MLP

Simple network showing DeepSDF style representation of a single shape. Uses NeRF style positional encoding for easier overfitting to shape.

Dependencies

This repository requires numpy, pytorch, pytorch-lightning and PyMarchingCubes (found here).

Training

Train using the following command from root directory:

python trainer/train_implicit.py

Expected Output

expected_output

File Structure

Folder Description
data/hollow_knight.npz Contains the raw data for training
dataset/point_sdf_pair.py Contains dataset class for training
model/implicit.py Implicit network specification
trainer/train_implicit.py Training module
outputs Mesh output dumped here while training
runs Checkpoints saved here while training

3D Model Used

Hollow knight (https://sketchfab.com/3d-models/hollow-knight-86c4d4bfc46645169211130e3dd2c810) by EduVelazquez is licensed under Creative Commons Attribution (httpcreativecommons.orglicensesby4.0).

About

Simple MLP for representing the SDF of a single shape

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published