lainy on Nostr: kaia i don't think the model is the problem. maybe take some example images of things ...
kaia (npub1c4s…aqv4) i don't think the model is the problem. maybe take some example images of things that you want and some example images of things that you don't want, then get an embedding from that and sort the images based on which type is closer to the embedding of the given image.
Published at
2024-07-12 10:45:55Event JSON
{
"id": "fc40bbbec78424ab169895cb2921663e272fd5af4494107636cbb69b7458b6a8",
"pubkey": "776ed1a547e2693a2c964e4824d6306a11aa364cd9c798f3e1ccd638af3d3725",
"created_at": 1720781155,
"kind": 1,
"tags": [
[
"p",
"c56086c6ba87cbcbfb72bcca534a1cf2ced76e95b7f57306367809fad85a616a",
"wss://relay.mostr.pub"
],
[
"e",
"01e1fab5129ff6fed625f7bf5cc6a224e6a72f0bdd22a726fc0ada0dfd5e8c23",
"wss://relay.mostr.pub",
"reply"
],
[
"proxy",
"https://lain.com/objects/f962333f-3076-44d0-b6d6-f0aaf50274a5",
"activitypub"
]
],
"content": "nostr:npub1c4sgd346sl9uh7mjhn99xjsu7t8dwm54kl6hxp3k0qyl4kz6v94qh5aqv4 i don't think the model is the problem. maybe take some example images of things that you want and some example images of things that you don't want, then get an embedding from that and sort the images based on which type is closer to the embedding of the given image.",
"sig": "fc1d66e6ade68eb2462497cb8f2b9590cc66c53a3bfc01f9193cfd5377827e4dee3cf2b35b5fa9024fd124a1a2f83ef8d603849b8131e36a2c02da84936eec33"
}
