Description

kushalkolar

opened on Aug 10, 2024

Issue body actions

I think the readme needs cleanup, ideas:

• update fancy examples gif, have some things from different domains. We also have axes now so something that demonstrates that! Maybe the 3D line stack animation example?

• Cleanup examples section, just link to the gallery and docs. Move the stuff about embedding in a Qt app to the docs along with examples.

• Add a few points from docs on what fastplotlib is #480 to quickly "advertise" near the top of the readme about what fastplotlib is and what it's meant for/what it can do.

• There might be a way to make the "supported frameworks" section simpler, some form of "write once, run anywhere" to get across same code on jupyter and Qt/glfw desktop apps. Move fast from prototyping to real web apps or desktop apps.

• A section that lists the developers?

Activity

1. kushalkolar commented on Oct 26, 2024

   

   kushalkolar

   on Oct 26, 2024

   MemberAuthor

   More actions

   Idea of "cool" visualizations for a gif in the readme, for presentations, and homepage of our eventual website, and examples gallery too if it doesn't already exist:

   1. Galaxy merger or a single galaxy, show lots of points, color based on something.
   2. Twisting polarization of a light wave, maybe we should make the vector graphic soon. One subplot with the wave, another a cylinder mesh showing light color with a cmap transform?
      • We could also add some other vector-field viz, along with a mesh too if we add that graphic soon?
   3. 3D calcium imaging viz, volume image in one subplot and just dots with changing size or color in another subplot.
   4. Pose tracking with skeleton of multiple animals in 3D, just lines and scatters for skeleton, and a line for each point showing trajectory.
   5. Streaming lots of line of ephys or other time series.
   6. Lorenz attractor.
   7. Exploration of a covariance matrix.
   8. Looking at the layers of a CNN as it's being trained, wait for wgpu and pygfx update modes refactor which will make this easier. Or see if pytorch has a way to do callbacks, like update some ImageGraphics data after every training epoch @clewis7 is there such a thing?
2. clewis7 commented on Nov 8, 2024

   

   clewis7

   on Nov 8, 2024

   Member

   More actions

   Looking at the layers of a CNN as it's being trained, wait for wgpu and pygfx update modes refactor which will make this easier. Or see if pytorch has a way to do callbacks, like update some ImageGraphics data after every training epoch @clewis7 is there such a thing?

   This is something we talked about yesterday after the talk, and I think I am going to start looking into it. I think the consensus was that pytorch might not readily expose where some of these things are being stored on the GPU so it might take some investigation.
3. clewis7 commented on Nov 8, 2024

   

   clewis7

   on Nov 8, 2024

   Member

   More actions

   Other ideas:

   • I am going to be updating the improv demo, not sure if we would want to include that (we are going to try to develop some cool viz moving forward with improv and SPLAT)
   • Daniella has done some super amazing work with flow fields and diffusion models, I am hoping to sit down with her and make a viz using imgui that allows you to watch the integration through time and change some hyperparameters (would be super cool)
4. clewis7 commented on Nov 8, 2024

   

   clewis7

   on Nov 8, 2024

   Member

   More actions

   I like the first item on the list as well. John agreed that astro would be a great place to maybe start doing some demos and getting in touch with people. I know you previously talked to astro people at flatiron before...
5. kushalkolar

   assigned

   clewis7

   on Jan 16, 2025
6. kushalkolar

   closed this as completedin #704on Jan 26, 2025

Issue actions