Faster Multi-Object Segmentation using Parallel Quadratic Pseudo-Boolean Optimization (supplementary material)

Supplementary material for the Faster Multi-Object Segmentation using Parallel Quadratic Pseudo-Boolean Optimization paper presented at ICCV 2021.

Content

This item contains code, notebooks and results for the experiments presented in the paper, including:

• C++ code and Python wrapper code for P-QPBO and M-QPBO algorithms. This is located in the shrdr Python package. See GitHub for the latest version of the package.
• Python code for building sparse layered graphs. This is included in the slgbuilder Python package, which can also be found on GitHub.
• Jupyter notebooks for re-creating the QPBO experiments and analyzing the results of both the QPBO and the non-QPBO experiments.
• CSV files with the experimental results presented in the paper.
• PDF with Proof of equivalent labeling for paper.

The item does not contain C++ code for the non-QPBO algorithms. Code for parallel BK and EIBFS algorithms can be found here. It also does not contain the raw image data of the nerves and nuclei used in the experiments. Please see the README.md, notebooks, or paper for links to the data.