基于Matejek等人的工作,对compresso算法进行了非正式打包。
compresso的Python项目详细描述
Compresso:用于连接组学的有效分段数据压缩
*注:这是Matejek等人作品的非官方包装,可在此处找到:https://github.com/VCG/compresso*
Recent advances in segmentation methods for connectomics and biomedical imaging produce very large datasets with labels that assign object classes to image pixels. The resulting label volumes are bigger than the raw image data and need compression for efficient storage and transfer. General-purpose compression methods are less effective because the label data consists of large low-frequency regions with structured boundaries unlike natural image data. We present Compresso, a new compression scheme for label data that outperforms existing approaches by using a sliding window to exploit redundancy across border regions in 2D and 3D. We compare our method to existing compression schemes and provide a detailed evaluation on eleven biomedical and image segmentation datasets. Our method provides a factor of 600-2200x compression for label volumes, with running times suitable for practice.
论文:Matejek等人,,“Compresso:用于连接组学的有效分割数据压缩”,医学图像计算和计算机辅助干预国际会议论文集,2017年10月14日。[CITE| PDF]
要求
- Python 3.5+
- 康达
设置
git clone https://github.com/vcg/compresso &&cd compresso conda create -n compresso_env --file requirements.txt -c chen -c sunpy -c conda-forge -c auto -c indygreg source activate compresso_env # for Compresso scheme as presented in MICCAIcd experiments/compression/compresso; python setup.py build_ext --inplace # to run the neuroglancer compression schemecd ../neuroglancer; python setup.py build_ext --inplace # for Compresso v2 that is under developmentcd ../../../src/python; python setup.py build_ext --inplace
压缩分段堆栈
此存储库中有两个版本的Compresso。在src文件夹下有一个更新的c++和python版本,它扩展了MICCAI中的Compresso方案。除其他外,该算法实现位压缩,以进一步改善压缩结果。在
experiments/compression/compresso
中的压缩方案与MICCAI论文完全一致。在
压缩分段堆栈
要对自己的数据测试Compresso,只需使用:
^{pr2}$实验
# the dataset must be in hdf5 format.
experiments/run.py COMPRESSO LZMA ac3 -r 1 -s 1 -d '/<PATH>/<TO>/<DATA>'
用法:
usage: run.py [-h] [--directory PATH] [--runs NUM] [--slices NUM]
[--verbose]
encoding compression dataset
positional arguments:
encoding name of encoding scheme
compression name of compression scheme
dataset name of data set
optional arguments:
-h, --help show this help message and exit
--directory PATH, -d PATH
path to data directory
--runs NUM, -r NUM number of runs (default: 1)
--slices NUM, -s NUM number of slices per dataset (default: -1 (all))
--verbose, -v print progress (default: False)
确保数据集位于~/compresso/data/
中或指定位置。论文中的数据可以在这里找到:
- AC3:http://www.openconnectomeproject.org/kasthuri11(kassuri等人,新皮层体积的饱和重建。单元格2015.)
- 克里米:http://www.cremi.org
- CYL:http://www.openconnectomeproject.org/kasthuri11(kassuri等人,新皮层体积的饱和重建。单元格2015.)
- SPL脑图谱:http://www.spl.harvard.edu/publications/item/view/2037(Halle M.,Talos I-F.,Jakab M.,Makris N.,Meier D.,Wald L.,Fischl B.,Kikinis R.基于多模态MRI的大脑图谱。2017年1月1日)
- SPL膝关节图谱:http://www.spl.harvard.edu/publications/item/view/2037(Richolt J.A.,Jakab M.,Kikinis R.SPL膝关节图谱。SPL 2015年9月)
- SPL腹部地图集:http://www.spl.harvard.edu/publications/item/view/1918(Talos I-F.,Jakab M.,Kikinis R.SPL腹部地图集。SPL 2015年9月)
- BSD500:https://www.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/resources.html(轮廓检测和分层图像分割P.Arbelaez,M.Maire,C.Fowlkes和J.Malik。IEEE TPAMI,第33卷,第5期,第898-916页,2011年5月)
- VOC2012:http://host.robots.ox.ac.uk/pascal/VOC/voc2012/(Everingham,M.和Van~Gool,L.和Williams,C.K.I.和Winn,J.和Zisserman,A.,PASCAL Visual Object Classes Challenge 2012(VOC2012)结果)
论文结果
压缩性能
联合使用Compresso和NeuroScarer的通用压缩方法的压缩比。Compresso与LZMA配对产生了所有connectomics数据集(左)的最佳压缩比,其他数据集(右)的平均压缩比(五分之四)。在
- 项目
标签: