Execution of Splitter Node on RGB Image takes 99% of the CPU and >4go RAM


In intro, I have to mentionne that I am a real newbie in Image Processing.

I want to detect the orange and yellow colors in some pictures by trainning a classifier.

To do so, I am trying to find the pre-processing that could improve my classifier. But when I tried to split my RGB image in the channel-dimension, the execution is very long (25% after 15 minutes) and consum an unusual amont of CPU/Ram. I try with only 4 images but same issue…

I surely did something wrong. If someone here know what I will be so grateful.

Thanks you very much in advance for your answers.

PS: I attached some pictures form my sample pics_sample.zip (76.6 KB)

Hi @AmauryD,
I think I found the reason for the slow processing, it lies in the way the column selection works in the Splitter node: You need to select the dimensions that are staying together, not the one you want to split. See the following screenshot for an example configuration, which will result in the channel dimension being split:

The way this dimension selection works is sadly not that intuitive, so your confusion is very understandable :slight_smile:. I hope you have lots of success with your processing and feel free to ask more questions if you get stuck.


Hi Gabriel,

Thank you for this quick answer ! Yes this way that’s work much better. And sorry my bad, I should had read the doc. more carefully.

Thanks again and good afternoon


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