I used Matlab for image processing for years. Tried to switch to Python 10 years ago but too many tools were still missing. Tried again 5 years ago and haven't touched Matlab ever since! The combination scikit-image + @ProjectJupyter was a real game-changer! A few more things:
Complex ML tools for image denoising like content-aware image restoration #CARE (github.com/csbdeep/csbdeep) or point-scanning super-resolution #PSSR (github.com/BPHO-Salk/PSSR) which are documented as Jupyter notebooks that really work "out of the box".
And it's really easy to share all this. For teaching you can setup a Jupyter environment using the super-easy to use "littlest JupyterHub" (tljh.jupyter.org). You can put your notebooks in an online repository and have ANYONE running it interactively with @mybinderteam
Finally, using tools from @QuantStack, turn interactive features of notebooks (#ipywidgets) into web-apps! E.g. for a course on very basic image processing I used the #voila gallery (github.com/voila-dashboar…) to create small interactive illustrations fl-7-45.zhdk.cloud.switch.ch
And of course one should never forget that this ecosystem only exists because of all the people who spend countless hours to maintain it (often for free, see e.g. github.blog/2019-06-28-mai…). As a user who mainly builds on top of these tools I am very grateful to them!
For sake of completeness, the thread was supposed to be a reaction to this tweet but it got lost by twitter dark magics apparently...
