Ilastik - understanding export data


I have been using ilastik to classify some objects which are cells, first by generating a positive pixel map, and then by classifying the objects into 3 groups. When I export my data, I get some things I understand (size in pixels etc) then some quite confusing information - like Variance of Intensity_0 Variance of Intensity_1 and Variance of Intensity_2… how do I know which of these 3 I might be interested in - or does it not matter as long as I look at the same one for each object?

This is so that I can plug my data into a PCA visualization tool, but also just for my own benefit


Hi @Ayman_Dsouza,

welcome to the community!

And great to hear that ilastik seems to be doing a satisfactory job for you :slight_smile:

The object features should really be documented in our docs… but so far the documentation for the individual features is only accessible in the ilastik gui, and also only if you hover over a feature name in the object feature selection applet.

This is of course not ideal.
Also some feature names (for the PCA of the object pixels, e.g.) are named confusingly, as this discussion on “radii” in the feature table showed.

For the variance of intensity the help suggests

Variance of the intensity distribution inside the object. For multi-channel data, this feature is computed channel-wise.

So I assume your image data has 3 channels? This means that Variance of Intensity_0 will be the variance of all pixel values for the object in the first channel, and so on.

Hope this helps (a bit)