This looks nice and useful. I was thinking for a long time about a graphics manipulation program that would put every step as a separate line with all parameters visible. Then you can change any parameter from earlier steps as you see fit. It makes the graphics editing process like a little interactive script. You could then easily parametrize it further and even do batch processing this way.
-b meaning batch mode, -p meaning send the value of main to the output. That'll load the process.ws workspace, set cell A1 to be the image "fred.jpg", set main (the point of execution) to be cell A10, and save it to fred2.jpg.
I made a workspace for medical image analysis (my old job). It finds lungs in a 4D PET-CT scan, then fits a two compartment model per voxel and generates a volume image from k2. It has just over 10,000 cells (!!), 70GB of images, and executes in about 2 minutes using around 1GB of ram.
I then used a small bash script to loop over my dataset of 2,000 scans computing an inflammation index. It worked surprisingly well, IMO.
Comments
This looks nice and useful. I was thinking for a long time about a graphics manipulation program that would put every step as a separate line with all parameters visible. Then you can change any parameter from earlier steps as you see fit. It makes the graphics editing process like a little interactive script. You could then easily parametrize it further and even do batch processing this way.
I will have to check this out.
Yes, nip2 has a no-GUI batch mode.
You can make a workspace interactively that does some task, then run it with something like:
nip2 -bp --set 'Workspaces.process.A1=Image_file "fred.jpg"' --set main=Workspaces.process.A10 -o fred2.jpg process.ws
-b meaning batch mode, -p meaning send the value of main to the output. That'll load the process.ws workspace, set cell A1 to be the image "fred.jpg", set main (the point of execution) to be cell A10, and save it to fred2.jpg.
I made a workspace for medical image analysis (my old job). It finds lungs in a 4D PET-CT scan, then fits a two compartment model per voxel and generates a volume image from k2. It has just over 10,000 cells (!!), 70GB of images, and executes in about 2 minutes using around 1GB of ram.
I then used a small bash script to loop over my dataset of 2,000 scans computing an inflammation index. It worked surprisingly well, IMO.