>developer
>nothing to do
>really nothing to do
>made a bot that detects if a person in a image is caucasian or african
>why the fuck i made this
>+80% accuracy
>don't know for what i will use this
>asking myself if this can be used to bait and trigger lefties
Does anyone have any ideas?
Other urls found in this thread:
Add script that renames to neger.jpg
Also, do you want to share it either on pastebin or GitHub?
>tumbr...jpg
rope yourself
make it 90%+ and you probably can sell that shit to police depts. in the US (maybe in Brazil too)
>implying i found in tumblr instead of other site that preserve the file name
panic
Results: 5/5 white
Can detect multiple faces too
She needs more training, i only trained for 200 epochs and 123 images kek
Works better on adult images
Sure you did, (((user)))
program it to say argentinians are white and you can trigger Sup Forums
...
do me
white
Am I white?
>no human face found
>+80% accuracy
That's kinda disappointing.
What approach are you using? CNN's?
Kek, it knows. SHUT IT DOWN!
CNN2D 3x3 trained with 200 epochs with ~100 images, i will try to train with more images tomorrow
Does he flare his nostrils to look more negro?
So this is a neural network right?
>CNN2D
I googled that and the first result was an issue from GitHub, on the repo of this thing:
Is that what you're using?
Also what's it like being a developer in Brazil? Well paid? And what sort of development do you do?
Yes, i am using the keras backend
And i am not a professional developer kek
Yeah, a special kind of neural networks used *mainly* for image processing
CNN = convolutional neural network
tl;dr - instead of simple math functions, each neuron / node represents a kernel
>And i am not a professional developer kek
Why not?
>16y
MODS MODS
Sell to police departments, but make it that every face they analyze is African
create a website and post it up there.
i will try to do this week
>made a bot
>Yes, i am using the keras backend
So, you did jackshit?
...
fuggg, and here come they come... Good luck
>implying use a backend is just do three lines and the code works
I had to write all the code to make the network work with the data, make a crawler to get images, and also to develop the best network structure to work with the most accuracy
I don't have time to make pure derivatives, delta rule, and backpropagation from scratch, any developer use modules to not lost time reinventing the wheel, Keras uses Theano or TensorFlow, both uses Numpy, and so on
thx
Create a bot from scratch with a linear algebra module made from scratch in a language made from scratch, in a computer made from scratch, with transistors made from scratch with silicon/germanium found in your home and with bottle capacitors from scratch, and charcoal resistors from your charcoal mine and with the gold that your country stole from Brazil
share? it could be very useful for filtering porn
it could also be used for security systems. alert you if a nigger walked by your home camera
hey look a thread about nothing. but lets make sure all christian threads get nuked
Tomorrow i will try to train with more images and put in my GitHub
I think its better i share the entire code, my PC can't run many convolutions of large images, a Epoch with ~100 images of 64x64 takes at least ~6sec, 256x256 170sec
...
Christcux need not apply
Sounds cool although cvDetectFaces and comparing regional color averages would probably do at least as well
Make a Jew detector
Which language?
I would delete this post if I were you, else you risk a ban.
Tick the post in the top left, then click the "delete" button in the bottom right corner of the page.
oh shit i never noticed that. he totally fucking does that. This guy is the biggest crypto nigger i have ever fucking seen.
>crypto nigger
>How white is you're nigga test.