Code for the paper "Compact Model Representation for 3D Reconstruction"
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
3D_reconstruction
FFD
data
exp
external
graph_creation
python
show
utils
README.md
add_paths.m
data_paths.m
get_started.m

README.md

Compact Model Representation for 3D Reconstruction

Here you'll find the codes for the paper Compact Model Representation for 3D Reconstruction presented at 3DV 2017.

Datasets Used

In this work we used the ShapeNetCore.v1 and PASCAL3D+_release1.1 datasets.

Requirements

To create the embedding graphs you'll need the CVX library.

Getting Started with a Demo

Running a demo for the aeroplane category:

  1. Clone the repo:
git clone https://github.com/jhonykaesemodel/compact_3D_reconstruction.git
  1. Download Pascal3D+ dataset
  2. Open add_paths.m and change the path root with the path of the directory you placed the datasets (e.g. root = 'C:\datasets')
  3. Download the precomputed data (i.e. aeroplane embedding graph files, ShapeNetCore 3D model IDs for all 3D models used and its manually annotated 3D anchors) here
  4. Unzip it and place the directories ShapeNetAnchors, ShapeNetGraph and ShapeNetMat.v1 into the root directory (e.g. root = 'C:\datasets')
  5. Run get_started.m and have fun :)

Free-Form Deformation (FFD) Demo

To get some intuition about FFD, in the FFD directory you'll find demo_FFD.m, a tutorial FFD_tutorial.mlx, and also a simple FFD UI tool where you can select a control point and play with cursors to deform a 3D mesh model.

To run the FFD UI run FFD.mat and click Init to load the Standford bunny. Wait until the bunny and the FFD grid of control points are loaded before selecting the control points and playing with the cursors.

Reference

If you find this code useful in your research, please cite the paper:

@article{pontes2017compact3d,
  title={Compact Model Representation for 3D Reconstruction},
  author={Jhony K. Pontes and Chen Kong and Anders Eriksson and Clinton Fookes and Sridha Sridharan and Simon Lucey},
  journal={International Conference on 3D Vision (3DV)},
  year={2017}
}