Cannot convert binary image to label if applied on a cropped image

Dear KNIME experts,

I have recently created a workflow that aims to use the interactive annotator to manually select a rectangular region of interest. This region is used to crop the image and to subsequently process/analyse only the manually cropped region. However, when I try to apply the global thresholder and the "image to labeling" node, I failed to obtain the label. Interestingly this works on the uncropped image. I have no idea why the "image to labeling" cannot be applied on a cropped image. The problem does not appear to be related to different images sizes in the column: if I apply the workflow after filtering to one row, the problem persists.

Would be very grateful for any advise. Thanks in advance. An example workflow with embedded images is attached.



I'm not sure if I understand the problem: What exactly do you mean by "obtain the label"? I get more or less the same results in both parts of the workflow, just one is cropped? 

Two things which I found anyway:

(a) You can/should disable the option "Add unique segment id" in the interactive annotator, use this option only if you have multiple rectangles per image. If you have multiple rectangles per image I suggest to use the "Filter on segment labels" with "Seg*".

(b) Maybe instead of image to labeling you want to use the "Connected Component Analysis" node?!



Hello Christian,

thanks to your response I could solve the problem. It appears to be a memory issue. The binary image was correctly converted into a label, but this label was not shown when I opended the table view on my computer. Changing the memory to "write to disc" was solving this issue. After update from KNIME 3.1. to 3.2. the problem did not occur anymore, even if images were not written to disk.

great.  If you have further problems, let me know!!



This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.