search

JaneliaSciComp / multiscale_benchmarking

0 stars 0 forks source link

chunking of outputs #1

Open d-v-b opened 3 years ago

d-v-b commented 3 years ago

@trautmane what chunk size did you use for the outputs? I couldn't find a reference to the output chunking in the bash script but maybe I'm blind? or is it hard-coded in java somewhere?

trautmane commented 3 years ago

I don't specify an output chunk size, so Igor's down-sampler keeps the same chunk size as the source (which in this case is 128,128,64).

See https://github.com/saalfeldlab/n5-spark/blob/master/src/main/java/org/janelia/saalfeldlab/n5/spark/downsample/N5DownsamplerSpark.java#L341-L343

I thought it made sense to simplify the downsample as much as possible since Saalfeld wanted to focus on comparing infrastructure libraries.