Skip to content

Latest commit

 

History

History
113 lines (89 loc) · 5.02 KB

README.md

File metadata and controls

113 lines (89 loc) · 5.02 KB
Raw

Description

RigidMotionsMapleTools is a set of Maple modules for which the main objective is to generate unique neighborhood motion maps (NMM, for short). These neighborhood motion maps are the images of an image patch which is a finite set of integer points. More information about the alpha version of the algorithm were published in: Pluta, Moroz, Kenmochi, Romon,Quadric Arrangement in Classifying Rigid Motions of a 3D Digital Image, CASC16, Springer 2016. You can obtain a preprint from HAL for free.

Quick Install

  1. Download or clone the repository.
  2. cd <path to the cloned repository/scripts>
  3. ./Install.sh -d="<path to the cloned repository>"

Install

  1. Download or clone the repository.

  2. Edit (or create) Maple initialization file and add to it paths to the files (the order matters):

    1. RealAlgebraicNumber.mpl
    2. RigidMotionsParameterSpaceEvent.mpl
    3. RigidMotionsParameterSpaceRegister.mpl
    4. RigidMotionsParameterSpaceCommon.mpl
    5. RigidMotionsParameterSpaceDecompositionRecursive.mpl
    6. RigidMotionsParameterSpaceDecomposition.mpl
    7. RigidMotionsRecoverNMM.mpl

  3. Additionally, FGb library by Jean-Charles Faugère can be installed to improve generation of univariate polynomials. For more information see: http://www-polsys.lip6.fr/~jcf/FGb/index.html. We do recommend to install FGb library in version 1.61.

Quick Uninstall

  1. cd <path to the cloned repository/scripts>
  2. ./Install.sh -u

Examples

To compute a sorted list of events for 6-neighborhood (for documentation of the parameters see the source file):

with(RigidMotionsParameterSpaceDecompostion);
LaunchComputeEvents([a,b,c], "N1.db", "N1", [-1, 0, 1]);

To run or resume computations (for documentation of the parameters see the source file):

with(RigidMotionsParameterSpaceDecompostion);
LaunchComputeSamplePoints([a,b,c], "N1.db", "N1");

To compute neighborhood motion maps (for documentation of the parameters see the source file):

with(RigidMotionsRecoverNMM);
LaunchFindDistinctSamplePoints([a,b,c], "N1", [-1,0,1], "N1.db");
LaunchComputeNMM([a,b,c], "N1", [-1,0,1], "N1.db");

Submitting jobs to a cluster

The directory scripts/cluster contains several scripts meant to easy the process of running parts of the code in a cluster. The only part of the code which does not work in a cluster is LaunchComputeEvents. The only supported job subbmitting command is qsub.

The main script Build.sh creates a self-executable bash script which contains all the code and input data needed to run computation in a cluster. To build a self-executable script run:

./Build.sh -d=<database.db> -n=<no. nodes> -o=<output.shx> -m=<scpript.mpl>.

The input parameters are:

  • -d -- path to a database which has to contain at least computed events
  • -n -- number of nodes in the cluster
  • -o -- where to write the output self-executable script
  • -m -- a maple script which contains a set of instructions to be run on each node in the cluster (see an exemple of such a file: scripts/cluster/MapleScriptExample.mpl)

To submit computations to the cluster you have to execute the self-executable script obtained from scripts/cluster/Build.sh and provide as arguments

  • -d="<path/to/shared/dit>" -- path to a shared directoy accesible in the same way by all the nodes
  • -b="integer" -- an optional integer argument which stands for a beginning of a range of the input databases to be proceed.
  • -e="integer" -- an optional integer argument which stands for an end of a range of the input databases to be proceed.

Moreover, any argument given after the above arguments will be transfered to qsub. The self-executable script will create a folder <SHARED_DIR>/selfextract.XXXXXX/DB, where XXXXXX stands for some random string, the output is stored in a number (equal to the number of nodes provided to Build.sh via -n parameter) of database files. For the moment the files have to merged by hand but a script to facilitated this operation is planned.

Additional information

The computations are memory and time expensive. The implementation uses the Maple Grid framework to distribute (locally) the computation on a given number of nodes. While, the problem for 6-neighborhood can be solved on a desktop machine in a relatively short time, computations for bigger image patchs can take weeks even for a machine with dozens of cores and hundreds of gigabytes of memory. Note that, if the FGb library is installed then computations are relatively slower but less events are generated.

DOI