A nifty library for 2D and 3D image segmentation, graph based segmentation an opt. This library provided building blocks for segmentation algorithms and complex segmentation pipelines. The core is implemented in C++ but the suggested language to use this library from is python.

To use nifty one needs to checkout some submodules via:

git submodule init
git submodule update

If WITH_MP_LP is active, one needs:

git submodule update --init --recursive

A very tentative documentation of the nifty python module.

• Multicut:

  • Multicut-Ilp (Kappes et al. 2011)
    • Multicut-Ilp-Cplex
    • Multicut-Ilp-Gurobi
    • Multicut-Ilp-Glpk
  • Decomposing Solver (Alush and Goldberger 2012)
  • Cut Glue & Cut (Beier et al 2014)
    • Cut Glue & Cut - QPBO
  • Greedy Additive Clustering / Energy based Hierarchical Clustering (Beier et al. 2015)
  • Fusion Moves for Correlation clustering (Beier et al. 2015)
    • Perturbed Random Seed Watershed Proposal Generator
    • Perturbed Greedy Additive Clustering Proposal Generator
  • Kernighan-Lin Algorithm with Joins (Keuper et al 2015)
  • Message Passing for the Minimum Cost Multicut Problem (Swoboda 2016)

• Lifted Multicut: (Andres et al. 2015, Keuper et al 2015)

  • Kernighan-Lin Algorithm with Joins (Keuper et al 2015)
  • Greedy Additive Clustering (Keuper et al 2015)
  • Lifted-Multicut-Ilp (does not scale to meaningful problems, just for verification)
    • Lifted-Multicut-Ilp-Cplex
    • Lifted-Multicut-Ilp-Gurobi
    • Lifted-Multicut-Ilp-Glpk
  • Fusion Moves for Lifted Multicuts (Beier et al. 2016)
    • Perturbed Random Seed Watershed Proposal Generator
    • Perturbed Greedy Additive Clustering Proposal Generator
  • Message Passing for the Minimum Cost Multicut Problem (Swoboda 2016)

• Mincut/Maxcut:

  • QPBO

• Agglomerative Clustering

  • Easy to extend / Custom cluster policies
  • UCM Transform

• CGP 2D (Cartesian Grid Partitioning)

• Many Data Structures:

  • Union Find Data Structure
  • Histogram

• Coming Eventually:

  • MultiwayCut:
  • ModifiedMultiwayCut:
  • LiftedModifiedMultiwayCut:

The Python API is at present the easiest to use. The C++ API is mostly for power users. We recommend to use library from Python. Almost any class / function in the Python API is calling into C++ via pybind11.

The easiest way to install this version is via conda (right now it is not available with gurobi or cplex, you will need to build from source for this):

conda install -c conda-forge nifty

To build nifty from source, you will need to set up an environment with all necessary dependencies. We recommend to use conda for this as well and provide the file environment.yaml` for this. You can set up this environment and build nifty with the following commands:

$ conda env create -f environment.yaml
$ conda activate nifty-dev
$ mkdir bld
$ cd bld
$ cmake -DCMAKE_PREFIX_PATH=/path/to/conda/envs/nifty-dev -DWITH_Z5=ON -DWITH_HDF5=ON -DWITH_ZLIB=ON -DBUILD_NIFTY_PYTHON=ON ..
$ make

This would build nifty with additional hdf5 and z5 dependencies; and add zlib support for z5. You can find all build options in CMakeLists.txt; note that you may need to install additional dependencies.

TODO

TODO