search

erikbern / ann-benchmarks

Benchmarks of approximate nearest neighbor libraries in Python
http://ann-benchmarks.com
MIT License
4.99k stars 747 forks source link

Graphs hard to parse for color blind people #210

Open SeanPedersen opened 3 years ago

SeanPedersen commented 3 years ago

Hey very cool project! I find the performance graphs really hard to understand though because of my red-green color blindness. I can imagine many others with worse color blindness having even worse issues. Is there a textual comparison table available? This would help a lot.

Another option could be to make the diagrams interactive and highlight the current selected algorithms performance in some way.

erikbern commented 3 years ago

There's actually interactive graphs available here! http://ann-benchmarks.com/glove-100-angular_10_angular.html

I'm not colorblind but honestly I find it hard to parse too. Maybe we could limit it to a smaller number of algorithms

SeanPedersen commented 3 years ago

Cool looks like I made a good choice using https://github.com/nmslib/hnswlib/ for my project.

Yes maybe reduce to the top 10 and the interactive graph is awesome, even for me! I suggest linking to it more prominently in the top section of the README as I was not aware of it after spending some time looking for it.

lc0 commented 3 years ago

@erikbern any plans to use interactive plotting like Altair in python. Would help with more algorithms and still all pythonic without much overhead

erikbern commented 3 years ago

I'd be open to doing that. If you have time, feel free to give it a shot and submit a pull request!