Newswitches

fusion takes to long with blending and full resolution

Dualchannel ... sorry that it took so long

compute_on is not present in Snakefile

compute_on, directory_cuda and cudafourierconvolution System.getProperty
now correct

For registration:
New added downsampling
New added Difference of Gaussian Detection
Difference of Gaussian with GPU or CPU

Switching between Difference of Mean and Gaussian is possible

Switching between GPU and CPU deconvolution is possible

Allows to process dual channel datasets were only one channel
contains beads

Added a switch in .yaml file for selecting dublication of transformation
Added a part in the snakefile for dublication of the transformation
in the pipeline

Update to newest master

New bsh scripts .. Thanks!

BUG: "Single Channel" is not a valid choice for "Process_channel"
When choosing Single channel in reg_process_channel then
this is not a valid input for the macro.
Now reg_process_channel is set to All Channel for Single
Channel processing

External transformation has a bug it breaks the .xml file
External tranformation is disabled for now.

Added new cluster settings for resave_hdf5_output in cluster.json

BUG in export_output.bsh: took subsampling and chunk sizes from
resave hdf5. But these were not good.
Added new chunck size and subsampling option

Started to update fusion

Added dual channel processing for these parts
updated .bsh file

Added channel selection

Merge branch 'yaml-support' of https://github.com/psteinb/snakemake-workflows into psteinb-yaml-support

Conflicts:
	spim_registration/timelapse/Snakefile

more bugfixes related to removing superfluous information for .tif files

used a regexp replace to look for the correct input files in define_xml_tif (just looking for *.tif might not do it as the deconvolution might produce .tif as well)