-
Notifications
You must be signed in to change notification settings - Fork 2
/
TARBALL_README.txt
139 lines (89 loc) · 4.89 KB
/
TARBALL_README.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
Obtaining and linking CPLEX:
============================
If you have CPLEX already installed on your machine, you can link to it to via calling the
install-cplex-shared-libs.sh script:
./install-cplex-shared-libs.sh /path/to/cplex-dir /path/to/multicut-software
If you don't have CPLEX installed yet, you must apply for an academic license, for
details see the section Download IBM CPLEX under
http://ilastik.org/documentation/basics/installation.html
For OSX, please use the "OSX" CPLEX installer, not the "darwin" version.
Obtaining the data:
===================
We have bundled the relevant data for reproducing the SNEMI and ISBI experiments, s.t.
they can be directly used with the scripts provided (see below):
http://files.ilastik.org/NaturePaperDataUpl.zip
After unpacking the data, you should have a folder "NaturePaperDataUpl" with two subfolders
for the experiment data, we will refer to as "\path\to\NaturePaperDataUpl" below.
Running the scripts:
====================
After unpacking the executables, the directory <multicut_root>
(where this README resides) will contain three scripts for the
different use-cases of the software:
- run_mc_2d.sh: Reproduce the ISBI experiments or similar.
- run_mc_anisotropic.sh: Reproduce the SNEMI experiments or similar.
- run_mc_isotropic.sh: Segment isotropic volume data.
run_mc_2d.sh
------------
You can use this script for anisotropic data with groundtruth containing
membrane vs. non-membrane labeling, as in the ISBI 2012 challenge.
Usage:
./run_mc_2d.sh \path\to\NaturePaperDataUpl\ISBI2012 \path\to\ISBI_output_folder
Make sure that the output folder already exists. After a successful run, it
will contain the files 'multicut_segmentation.tif' (a label map stored as
multi-page TIFF) and 'multicut_labeling.tif' (a boundary map, stored as
multi-page TIFF).
Optional argument:
--use_lifted=True
Use the lifted multicut instead of the standard multicut. This option
improves the results, but is deactivated by default, because it takes up
more RAM and runs considerably longer.
run_mc_anisotropic.sh:
----------------------
Use this script for anisotropic data with instance level groundtruth,
as in the SNEMI3D challenge. Note, that reproducing the SNEMI experiments
takes several hours.
Usage:
./run_mc_anisotropic.sh \path\to\NaturePaperDataUpl\SNEMI3D \path\to\SNEMI_output_folder
Make sure that the output folder already exists.
Optional arguments:
--use_lifted=True
Use the lifted multicut instead of the standard multicut. This option
improves the results, but is deactivated by default, because it takes up
more RAM and runs considerably longer.
--snemi_mode=True
If set to True, runs experiments with the same settings as our SNEMI3D experiments.
In particular, this uses the precomputed oversegmentation that was corrected for myelin.
For testing purposes, you should keep this set to False (the default). The True setting
may take longer, but reproduces the publication's results for the SNEMI dataset.
run_mc_isotropic.sh
-------------------
Use this script for isotropic volume data with instance level groundtruth.
Note that we do not provide exemplary data in 'NaturePaperDataUpl.zip' for this
use-case.
Usage:
./run_mc_isotropic.sh \path\to\your_data \path\to\output_folder
The input data folder must be structured like the SNEMI3D folder provided by us.
Make sure that the output folder already exists.
Optional arguments:
--use_lifted=True
Use the lifted multicut instead of the standard multicut. This option
improves the results, but is deactivated by default, because it takes up
more RAM and runs considerably longer.
For running any of the scripts above on your own data, you need to store it in the same format
and under the same names as the data in the exemplary folder.
-------------------------------------------------------------------------------------------------------------
To generate probability maps with ICv1, please refer to the relevant documentation in /software/nnet
Development and Packaging
=========================
This software was packaged with the conda package manager.
All dependencies can be downloaded from anaconda.org, with some packages sourced
from the 'ilastik' conda channel.
To install the multicut software (and dependencies) to your own conda environment, use this command:
conda install --name my-env --channel ilastik multicut_pipeline
In fact, this entire tarball can be reproduced with the following commands:
$ conda create --name multicut-software --channel ilastik multicut_pipeline
$ tar -czf multicut-software.tar.gz $(conda info --root)/envs/multicut-software
For build recipes of individual packages, refer to the source repository:
https://github.com/ilastik/nature_methods_multicut_pipeline
Note: Development is recommended on Mac/Linux platforms only, using conda-build v2.
Reproducing these binary pacakges on Windows requires a special version of conda-build.