label-fusion: Volumetric Fusion of Multiple Semantic Labels

C++ code to fuse multiple object labels or mask into OctoMap, which can be then used for 3d reconstruction of objects. It works with and without depth inputs, so can be applied for depth insensible objects: texture-less (for stereo), black (for ir), and transparent.

Requirements

• OpenCV (tested with OpenCV 2.4.8)
• Eigen
• octomap (modified version) (automatically resolved by build.sh)
• PCL (tested with PCL 1.7.1)

Installation

git clone https://github.com/wkentaro/label-fusion.git
cd label-fusion
./build.sh

Demo

Fusion of multiple labels

./demo.py label_view  # see inputs
./demo.py label_fusion
./demo.py label_fusion --depth  # use depth

Fusion of multiple masks

We also support fusing multiple masks:

./demo.py mask_view  # see inputs
./demo.py mask_fusion
./demo.py mask_fusion --depth  # use depth

License

MIT License (see LICENSE file).