Installation via pip install

I’m trying to install through Anaconda and am trying the simple "pip install histomicstk --find-links https://girder.github.io/large_image_wheels." This is my result:

Any thoughts?

What OS are you running? We only have wheels published for linux, so if you are using OSX or Windows, I wouldn’t expect this to work. If you are on a linux of any sort, I would only expect this error if your system can’t reach pypi where histomicstk is published.

  • David

That makes sense. I was trying to run on Mac OS. It is my understanding, then, that I install large_image as a python package and proceed to download HistomicsTK through:

$ git clone https://github.com/DigitalSlideArchive/HistomicsTK.git
$ cd HistomicsTK
$ pip install -e .

The issue is that when I clone large_image and install the dependencies, this error pops up:

Any idea? I can’t seem to download an updated mapnik either

Currently, we list large-image[sources"] as a dependency in HistomicsTK. This installs all the available tile sources, but each of those requires the appropriate libraries. On linux, we have wheels that satisfy all of these. I know other OSX users have changed setup.py so instead of requiring large-image[sources], it just requires large-image. You’ll also need to install at whatever tile sources are needed for your image formats. For instance, if you are using svs files, you’ll need to use homebrew to install openslide, then install large-image-source-openslide.

I’ll make a change to HistomicsTK’s setup.py file to not install tile sources on non-linux systems, which at least will remove the need to change setup.py.

I’d love to use HistomicsTK’s color normalization and related facilities on my Intel MacBook Pro, but of course by Apple’s making clang the default for all C-related compilers and other things installation via

$ git clone GitHub - DigitalSlideArchive/HistomicsTK: A Python toolkit for pathology image analysis algorithms.
$ cd HistomicsTK
$ python3 -m pip install -e .

is full of errors. Thanks for changing HistomicsTK’s setup.py file, but it still looks like a non-trivial time sink to get HistomicsTK installed on a Mac. Perhaps you could point me to a friendly user that has succeeded?

If not, I’ll try to use use the techniques in “Ensemble Prostate Tumor Classification in H&E Whole Slide Imaging via Stain Normalization and Cell Density Estimation” by Weingnat et. al. which seem pretty straightforward (before actually trying them).

Cheers,
Jon Sauer