Emil Melnikov

Hi @goanpeca. Sorry for the late reply. I'm a bit unsure about adding this package to conda-forge: it's probably of little value to people in this state. I've pushed it...

> it's still super weird that there is not a scrollbar on your screenshot... I tried to make the ilastik window smaller on Big Sur, and the scrollbar briefly appears...

Steps to reproduce: 1. Create a new _Pixel Classification_ project 2. In the _Input Data_ applet, create 2 lanes by adding 2 _Raw Data_ images (I used `2d_cells_apoptotic.tif` for both)...

The out-of-bounds error happens because [subslots are saved in a data structure](https://github.com/ilastik/ilastik/blob/d2390fee59acbc3136d8ec15537bf800b9110030/lazyflow/operators/opConcatenateFeatureMatrices.py#L38). However, the core problem stems from the lazyflow design: data selection keeps lanes as `OpMultiLaneDataSelectionGroup`, which maintains lane...

> The two ideas for _annotations_ that we currently have floating around are > > 1. the ilastik ones, where we have axis aligned, anchored (by roi) vigra arrays that...

I think saving the cached probabilities is not that bad: PC+OC workflow should not be used on big datasets anyway, so increased project file size shouldn't be a big problem....

I vote for 2, but disable glob expansion by setting a special environment variable instead of escaping and/or passing `skip_deglobbing` parameters around. This is not pretty, but it is very...

> I think we should pass a normalized name. Instead of "Pixel Classificaton" it should be "pixel_classification" I also like normalized names, but that's not how ilastik names workflows internally,...

> right now, this will only show up in the pixel classification workflow. So a followup-pr should make this available in object classification, too. I think I'll just another commit...

> > I think we should pass a normalized name. Instead of "Pixel Classificaton" it should be "pixel_classification" > > I also like normalized names, but that's not how ilastik...