I have a Neural Network for image classification all set up and my 1080Ti is screaming for some more work. Let's me hear your ideas for funny/meme tier stuff to train it to recognize. I've been thinking about training it to recognize buttholes, but I think that's already been done. I was also thinking about training it to identify reaction images so I can clean up my folder of 10 years worth of dank memes, but I don't know.
I have not time for this, but you should acquire as many pepe examples as possible, and train a GAN on Pepes. Then, you can navigate the latent space to generate new pepes. Try it, you'll learn a lot as well as give us new material. Get on it.
Caleb Roberts
parse porn video thumbnails and give the appropriate tags
Evan Thompson
Train it to distinguish traps from actual women
Gavin Hall
>Recognise wether its a dick or a pussy
Good idea.
I just had a better idea, what about recognizing traps? It can be hard to tell some times, but the facial features can often have nuances that a neural network might pick up on.
I also wonder if there are physical cues that might make it possible to detect gays? Before you react, hear me out. I don't think there is, however have it been studied? Maybe lesbians have higher testosterone levels or some shit which alters their facial composition? Would be interesting to try out.
I'm also thinking about training one to spot black people among pictures of mixed black people and gorillas.
Easton Anderson
>Train it to distinguish traps from actual women
Good idea, I just had the same one:
Grayson Taylor
What kind of tags?
Actually, classifying porn is a pretty good idea. I could try to detect things like vaginal sex, anal, oral, number of people in the scene, who have sex with who and what acts they perform, black, white, asian, lication (beach, indoor, outdoor), and other parameters. Also, known porn actors through facial detection and recognition.
Isaac Stewart
On my research to this I've found a guy who had tags for pretty much every possible position out there. Reading the names he came up for was pretty funny. However, all his gains are lost. I just found cached remainders.
Hunter Nguyen
yes yes yes yes yes
Jacob Jackson
Yes. And the output should be like this: >Trap -> Gay >Not Trap -> Not Gay
Tyler Brown
Got any links?
In my """"research"""" over the years, I've realized that fetish communities often have their own lingo/words. I suppose if this could be quantified as well, it would be a pretty popular search engine for porn. Basically type in your fetish and åerhaps some other keywords you want included and it will find you every porn video there is that matches based on the actual content in the videos.
Nolan Cook
Jews detection, traps detection Use Instagram or twitter to search dataset.
Automatic woman ratings. Input pic of hot female -> 8-10 Input pic of fat whale -> 1-3 (or whatever) Etc. Go trigger the feminists, user. Might take a while to build the dataset though
Xavier Miller
As in make it output from 1 to 10 depending on how hot the woman in the pic is
Ethan Cook
Indexxx get in thousands hot nude models, Instagram too and feminist blogs for other type classification.
Jace Scott
Begin very easy, hip-waist, legs as body, in face jaw,forehead or eyes lines.
Mason Flores
this, plus export the tech to Jow Forums
Jace Rogers
This is interesting, but how is hotness defined between us guys?
Pic related is obviously hot, but is it a 10/10? More to the point, if I were to label pictures of women, surely there would be too much inaccuracy. How do I judge if a given woman is a 5, 6 or 7?
I do know that if you average faces you get results that most people would consider 10/10, so perhaps I could instead compare how far from such an average face any given picture of a womans face is? The more delta, the lower the score?
Well mark zuckeberg work over it, made some page as hot or not? Using Instagram photos.
And data mine this babes rater com
Luis Gray
seconding the pepe generator
Isaiah Flores
What about looking at porn video watch count as a function of time? The hotter the porn star, the more popular she will be? Same with instagram photos, the hotter the women the more likes/subscribers/comments/etc ? Seems like a flaky and not very robust approach though
Sebastian Cruz
I already have a very extensive collection of both common and very rare pepes. I will look into it.
Daniel Anderson
Porn addict begin in two groups monster boobs and Lolitas and usually popular porn star begin for made hunders or thousand video in most brutal porn fetish.
Instagram is very tricky some semi famous girl will get thousands and some girls just upload semi nude photo.
Christopher Campbell
Holy grail approach is build GAN using beautiful women pics and made some website or hunders Instagram account to weight ... same approach to boobs,ass and body.
Using IA to build perfect human faces and body’s, IA porn apocalypse.
Cameron Harris
>perfect human faces perfect female faces. then you make an app that compares faces and rates yours. roasties will kill themselves en masse
Gabriel Scott
Gays usually have gay eyes as circular as that reasoning sounds, though this may inherently select for "faggot" gays and otherwise straight acting men who like dick might not actually share the same features. I'm not expert and my theories may just be lazily nigger rigged together intuition based arguments, but there does seem to be at the the minimum two distinct type of gays.
Matthew Gonzalez
I have about 16gb of pictures that are almost completely unsorted. Around 16000 images. I would love a way to automatically sort images by character depicted. And possibly create symlinks if the images has multiple characters. I originally tries to brute force this by using the e621 and Danbooru APIs to look up MD5 hashes and gather information, but a not insignificant portion of the images have MD5 hashes that are not in the database of either site.
the app should point out why the/which part of the face sucks
Oliver Martin
This. Let's see how long roasties will last when they get rejected by our AI overlords.
Ayden Wright
What kind of characters are in it? You could make a list of everyone, then scrape images from google for each one and create classes from them. Then you just train the network on them and hey presto, you can identify them in images. You can have 1000+ different classes.
Nathaniel Bailey
Actually not a bad idea.. Could even suggest ways to improve it (Loose weight, use this or this cosmetic to remove this spot or highlight this part, see a plastic surgeon, etc)
I think it would make for a popular app..
Luke Phillips
call it madonna
Lucas Gomez
or primadonna but I dunno how americans understand this word. it should have a positive connotation eventhough there are malicious intentions
Lucas Foster
Good ideas so far guys, I have a lot to try out. Still looking for something really creative though. Maybe I could replace eyes in pictures with buttholes? Would that have value?