These are chat archives for thunder-project/thunder

Nov 2015
Aaron Kerlin
Nov 19 2015 14:01
@jwittenbach A delayed thanks. I ran into this problem again today and your workaround did fix it.
Jason Wittenbach
Nov 19 2015 15:11
@aaronkerlin and @26tarun glad those fixes worked! :smile:
Nov 19 2015 21:42
@freeman-lab I installed thunder 0.6.0 from github and loading the binaries and all examples work like a charm on my University Cluster. Thanks for helping me figure that out.
Nov 19 2015 22:24

As I have moved on to tutorials and am now working with my data, I am running into some issues loading the image files.

My data are 16-bit multipage-tif stacks that are 1024x1024 with 2047 time points within the single tif. As per some of the above discussion with 16-bit images being loaded in thunder, I have tried a few things. These include loading the images with PIL on my cluster environment and converting the image stack to 8-bit. PIL can open both the 16 and 8 bit images fine. I am trying to load the files with:

data = tsc.loadImages(imagepath, inputFormat='tif')
data = tsc.loadImages(imagepath, inputFormat='tif', startIdx=0, stopIdx=10)

In both cases, I get:

nrecords: 1
dtype: None (inspect to compute)
shape: None (inspect to compute)
dims: None (inspect to compute)

When I try to inspect the structure, I get an overload issue and crashes and won't load.

Since I am running in interactive mode, I anticipate I may be overloading the executor's memory in trying to load these images. However, I hoped to fix that using the startIdx, stopIdx arguments.

Is there something I need to do differently to load a multipage tif stack correctly? Do startIdx and stopIdx work the same way in this setting?

tarun joshi
Nov 19 2015 22:56
This message was deleted
@jwittenbach : Thanks once again ....and... please guide us on a approach for our problem about batch feature extraction for a relatively unrelated Data set and indexing them?