/3dmodel-generator

Primary LanguagePython

3D-printable model generator

Project Description

This project is trying to use neural network and structural analysis to automatically generate new 3D printable models.

Currently we will use 3D-GAN which runs on GPU to generate 3D models and use other algorithms running on CPU to make generated model 3D printable.

Some models in ShapeNet after pre-processing.

Models generate by neural network

Project Source File

  • dataIO.py: data input and output model, and can transform mesh model to voxel model
  • setting.py: global setting variables
  • view.py: tools to visualize model or result
  • model.py: design neural network model to generate 3D models.
  • training.py: to train models generator use model designed at model.py. Use PyTorch to train 3D-GAN network to generate voxel models.
  • utils.py: some auxiliary functions, all other class will inherit this class

Environment Requirement

  1. nvidia drive, cuda 8.0+
  2. pytorch
  3. pathos: for multiprocessing, get from "pip install git+https://github.com/uqfoundation/dill.git@master" and "pip install git+https://github.com/uqfoundation/pathos.git@master"
  4. trimesh: to read model and transform it, get from "pip install trimesh"

ChangeLog

20180313

  • Add visualize process and will save module after every epoch's training

20180310 & 20180311

  • Fix some training problems.
  • Reorganizd project
  • Add main.py as interface to call other modules

20180309

  • Finished training procesure. First runable model

20180308

  • Finished model construction.

20180307

  • Finished view module. Can use mat files to generate model images

20180306

  • Finished dataIO module, can read and write model, can transform mesh model to voxel model and save them. Can randomly yield models.

Reference