Matt McCormick

This is related to specification of custom colormaps. We should resolve #186, then work on addressing this use case. Is there an example notebook we can use as a test...

@mrocklin good suggestion regarding `coarsen`! :+1: Some experiments are summarized here: https://gist.github.com/thewtex/079c59ea71a1da59bea5e14cb6b73d1f For large image support it looks like we should not convert NumPy arrays, ITK Images, etc. to Dask...

> In general I agree that the main advantage here is probably for larger-than-memory data. Yes, the prospect of eventually handling larger-than-memory with a distributed system is exciting :exclamation: >...

> It seems to me you have a 6.7 GB dataset you are working on. I feel your pain. ITK is going at about 4GB/s. You are shrinking by 10,...

> The reason this is important is scikit-image calls `np.pad` to pad the array before `downscale_local_means`. @lagru seems to have wrtiten `_fast_pad` for scikit-image which implements the algorithm above, but...

> you are downsampling 2048, 2048, 1600 by 10, 10, 10 which causes the array to get padded. The input array should be contiguous and its shape should be divisible...

@misterjoa thanks for the report, this is feature now on the roadmap :-).

Hi @ghosh034, does the notebook from `master` work for you? https://github.com/InsightSoftwareConsortium/itkwidgets/blob/master/examples/ImageJImgLib2.ipynb

@FuriouslyCurious @nickruof thanks for the report. Yes, we will updates for JupyterLab 3, but the installation experience for users should be much better.

Hi Robert! @haesleinhuepf A new alpha version is available for JupyterLab installable via: ``` pip install --upgrade --pre itkwidgets[lab] ``` with [docs on ReadTheDocs](https://itkwidgets.readthedocs.io/). Issues should be expected (alpha alpha),...