Usage:

mergeh5_chunks(configfile)

Inputs:

configfile: txt file that has all the parameters used in conversion

Fields:

inputfolder: input render folder that has the h5 octree
outname: fullpath of output h5
seqtemp: temporary file used to cache input tile paths that are found recursively in inputfolder at a certain depth. Depth is set to leaf node by default. useful to rerun conversion with different setting.
h5channel: color channel used in wrapper
ext: post-tag for naming each channel. we used 0/1 convention for our datasets.
maskThr: threshold to make data sparse, useful with compression
viz: 0 by default. for debugging purposes.
transform.txt arguments: [ox,oy,ox,sz,sy,sz,nl]: copied from transformed.txt. @TODO@ will get rid of this in the next version as we already pass inputfolder.
numCPU: number of cores used for paralel read. Useful to set a minumum number. Internaly we check max hardware cores, i.e. min(numCPU,maximum available CPU)

Sample config file:

# parameters

# render output to be merged
inputfolder = '/nrs/mouselight/SAMPLES/2018-10-01-prob'

# output h5 name
outname = '/scratch/classifierOutputs/2018-10-01/20181001_prob0/20181001_prob0'
# filelist sequence
seqtemp = '/scratch/classifierOutputs/2018-10-01/20181001_prob0/20181001_prob0-seq0.txt'
# copy scratch to /nrs/mouselight/cluster/classifierOutputs/2018-08-01/20180801_prob0

h5channel = 'prob0'
ext = '0.h5'

maskThr = 6500
viz = 0;

# from transform.txt of the render file
ox: 69445196
oy: 12917408
oz: 30199640
sx: 16173.689102564102
sy: 16073.159226190477
sz: 64319.413043478264
nl: 7

numCPU = 40