i have a collection of about 30,000 images that i have a script tweeting out every ten mins. the images are generally ‘aesthetically pleasing’ but theres a bunch of random images in there thats making me have to moderate the tweets, so this process is less autonomous than id like.

is there a way using TS or otherwise to feed a model say ~100 images that i handpick and it be able to tell me how ‘alike’ the image is to the other images? it seems hard bc im not classifying nude images (for example) so training the model would be difficult, right? im classifying a bunch of different things – fashion, landscapes, plants, etc. is determining if an image is ‘aesthetic’ too abstract for ML? If you have any resources please let me know! thanks!