No description or website provided.
C++ Other
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.
Input
TRW_S
cv_utils
BSplineSurface.cpp
BSplineSurface.h
CMakeLists.txt
ConcaveHullFinder.cpp
ConcaveHullFinder.h
DataStructure.h
LayerDepthRepresenter.cpp
LayerDepthRepresenter.h
ProposalDesigner.cpp
ProposalDesigner.h
README.md
Segment.cpp
Segment.h
StructureFinder.cpp
StructureFinder.h
TRWSFusion.cpp
TRWSFusion.h
main.cpp
utils.cpp
utils.h

README.md

Layered Scene Decomposition via the Occlusion-CRF

By Chen Liu, Yasutaka Furukawa, and Pushmeet Kohli

Introduction

This paper proposes a novel layered depth map representation and its inference algorithm which is able to infer invisible surfaces behind occlusions. To learn more, please see our CVPR 2016 paper or visit our project website

This code implements the algorithm described in our paper in C++.

Requirements

  1. OpenCV
  2. PCL
  3. gflags

Usages

To compile the program:

  1. mkdir build
  2. cd build
  3. cmake ..
  4. make

To run the program on your own data:

./LayeredSceneDecomposition --image_path="your image path" --point_cloud_path="your point cloud path" --result_folder="where you want to save results" --cache_folder="where you want to save cache"

To run the program on the demo data:

./LayeredSceneDecomposition --image_path=../Input/image_01.txt --point_cloud_path=../Input/point_cloud_01.txt --result_folder=../Result --cache_folder=../Cache

Point cloud file format:

The point cloud file stores a 3D point cloud, each of which corresponds to one image pixel. The number in the first row equals to image_width * image_height. Then, each row stores 3D coordinates for a point which corresponds to a pixel (indexed by y * image_width + x).

Contact

If you have any questions, please contact me at chenliu@wustl.edu.