3DRecon Toolkit

3D reconstruction toolkit, including a complete pipeline to run various 3D vision algorithms, and generate graphs for evaluation purposes. For more details, please visit the official website.

For WACV reviewers

All information linked to my identity have been removed from the official website. I also removed personal information on my Github account. Please stay within the official website, otherwise you could come across other Github sites that could reveal my indentity.

Dependencies

• MATLAB
• eigen3-nnls

Download

Clone this repository and download data.

git clone https://github.com/imkaywu/3DReconKit
cd 3DReconKit

Dataset

The real-world dataset can be downloaded in the Dataset page of the official website.

The synthetic dataset can be generated using the blender projects and scripts.

Algorithms

Algo Class	Algo	Summary	Source code
MVS	PMVS	Patch based Multi-View Stereo	PMVS
PS	EPS	Example-base Photometric Stereo	PSKit
PS	LLS-PS	Least squares Photometric Stereo	PSKit
SL	GSL	Gray-coded Structured Light	SLKit
VH	VH	Volumetric Visual Hull

Demos

1. Run 3D reconstruction algorithms on synthtic dataset to discover the effective properties, run

eval/synth/eval_prop/run.m
eval/synth/eval_prop/evaluate.m

2. Run 3D reconstruction algorithms on synthtic dataset to discover the mapping between problem conditions and algorithms, run

eval/synth/eval_algo/run.m
eval/synth/eval_algo/evaluate.m

3. Run 3D reconstruction algorithms on synthtic dataset to evaluate the performance of interpreter, run

eval/synth/eval_interp/run.m

4. Run 3D reconstruction algorithms on real-world datasets to evaluate the performance of interpreter, run

eval/real_world/run.m

Problem conditions

We are interested to find out the problem conditions under which an algorithm performs well. The process is divided into three step:

1. eval_prop: study the main effect of each property or interaction effect of each pair of properties;
2. eval_algo: evaluate the performance of each algorithm under a variety of problem conditions;
3. eval_interp: evaluate the proof of concept interpreter in terms of its ability to translate a user-specified description to a successful solution.

License

MIT