Conversation

Notices

1. Embed this notice
   James Endicott (o76923@kitty.social)'s status on Monday, 17-Jun-2024 05:02:56 JST James Endicott

   That's it. I'm adding an imagehash field to my master emoji file. Same name wouldn't work for consolidating because of how much name variance there is. Trying to match image files exactly isn't going to work either because of compression artifacts. Time to start reading up on how that works.

   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Monday, 17-Jun-2024 05:02:55 JST :blobcathug:

     in reply to
     @o76923 imagehash is far too imprecise
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Monday, 17-Jun-2024 05:16:15 JST :blobcathug:

     in reply to
     @o76923 i had the idea of a emoji compare software a longer time ago. Im not sure but depending on what you want this might not fit since it want to calculate a distance between two images... i used https://github.com/myint/perceptualdiff and i got, well not great results but far better results than some other stuff i found
   • Embed this notice
     James Endicott (o76923@kitty.social)'s status on Monday, 17-Jun-2024 05:16:22 JST James Endicott

     in reply to

     • :blobcathug:

     @Jain@blob.cat
     
     I'm certainly open to suggestions on what might work better. The colorhash algorithm that I've been fiddling with seems to be doing an okay job so far. Or, at least, one of the algorithms in the package does. The rest are not great for this specific task.

   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Monday, 17-Jun-2024 05:17:56 JST :blobcathug:

     in reply to

     • :blobcathug:
     @o76923 i somewhere got another sci paper which described better algorithms but i never implemented something