Furthermore, I have another question. I would like to use it to segment and classify cars on aerial images taken by a drone because for the moment I am trying to develop a workflow like this:
Hi,
That might be. Please check the data for missing values or empty tables, as those are most often the cause of workflow problems.
Kind regards,
Alexander
Hello,
the column NAME is completely empty. For this reason I got “duplicate row ID” error.
It is only in the part of the workflow after PYTHON SCRIPT node. How can I deal with it?
Hi,
I think I found the problem: in the String Manipulation node " extract ID from URL part 2" the expression is only suitable for file paths with forward slashes, but I suspect you are on Windows, where backward slashes are used. Please change the expression to this:
Hi,
the problem for the empty column is solved. Thanks for your big help!!! But now, it gives me another error: ERROR RowID 0:15:12 Execute failed: Encountered duplicate row ID “file:/C:/Users/davpr/Desktop/knime_4”
All the rows are called with this name and for this reason the name column cannot replace the RowID column.
Sorry to bother you!
Hi,
I think something is still wrong with the workflow. All row ids should be numbers (first one is 9004242). Unfortunately I don’t have a Windows PC at hand and cannot try it. Maybe @DaveK can help? He is the creator of that workflow.
Kind regards,
Alexander
I just tried executing the WF on Windows, which worked fine for. You should not have to do any modifications to the expression. Could you maybe post a screenshot of the output of the “Fix image dimensions” metanode?
thanks, that is what I meant. I found the problem. The Extract ID metanode aims to extract the folder name containing the image id from the URL column. For this it tries to chop away the file ending (everything after the last dot in the URL, i.e. .regions.txt). For this it searches for the position of the first dot. Now, your URL contains some additional dots because you put your workspace folder into your KNIME directory containing the version number in the folder name. This can be easily fixed using the lastIndexOfChar function when looking for the dot.
Sorry, for the inconvenience. This is a small bug on our side. I will update the WF (I will post a little notice here after I updated the WF). Either you can download the workflow again later, or you could try to fix it on your own as a little exercise.
I really want to thank you for your helpfulness. I changed the “extract ID from URL
part 1” node with “substr($URL$, 0, indexOf($URL$, “.r”))”.
I don’t know if it is a good way to solve the problem, but it worked.
For file size reasons, the example does not include an already trained model. You have to execute the Training branch first. The error on the DL Python Network Creator indicates problems with your Python environment. To do so, please follow this guide.
I do not know how your base environment is set up. Maybe there are some version conflicts. We strongly recommend creating a separate environment for the Deep Learning Integration. This way you can be sure that the installed versions are compatible with KNIME. You can simply do so by clicking the New environment.... Then chose either the CPU or GPU one, depending on if you have a compatible GPU. If you do not know, choose CPU.
I tried to create a new environment, and I got another error.
ERROR DL Python Network Creator 0:16 Execute failed: The external Python process crashed for unknown reasons while KNIME set up the Python environment. See log for details.
Unfortunately, I did not encounter this error before. Could you maybe upload your full KNIME log file here? (documentation where to find the log file)
Some ideas that come to mind:
If you use a GPU env, could you try again with the CPU one for testing reasons?
The error indicates a timeout. Did you try to executing the node several times?