What A Neural Network Thinks Of Your Selfies

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
You have access to a state-of-the-art neural network and you use it for this. Great idea! :rolleyes:

In this fun experiment we're going to do just that: We'll take a powerful, 140-million-parameter state-of-the-art Convolutional Neural Network, feed it 2 million selfies from the internet, and train it to classify good selfies from bad ones. Just because it's easy and because we can. And in the process we might learn how to take better selfies.
 
Interesting article...I wonder if the writer knows some principles of photography. He seems to ask a few odd questions as to why people do certain things in selfies which seem to be directly correlated to those rules.

Also, in the top 100 selfie pics he listed, at the bottom row center is King Joffrey being poisoned from the scene in Game of Thrones. Lol. I wonder how many thousands of people had to tag that pic with #selfie in Twitter to make that happen.
 
Forcing a state-of-the-art neural network to analyze millions of selfies?

When the machines turn on us, at least now we will know why.
 
I hope it decides that there is no such thing as a good selfie, because selfies are misguided pointless crap.
 
It looks like it was designed to detect cleavage. In the list of Top100 (which was all women), it's always face at top, boobs at bottom.
 
I like their optimal cropping algorithm where it cropped one girl right out of the image in favor of some traffic!
 
Well it's basic training was based on "likes" so yeah, hot chicks win. This is interesting from a technical perspective, but it's not based on a robots unbiased good / bad. It's a robot being taught to judge like the masses. So the computer can see an image that's way too up close and judge that it's probably not going to be well liked. Or it'll see that it has been framed well and will be more liked.
 
Anarchist4000 said:
It looks like it was designed to detect cleavage. In the list of Top100 (which was all women), it's always face at top, boobs at bottom.
Click to expand...

The program only rated things based on the number of likes. Humans were the ones giving boob pics more likes, not the machine. :D The program wasn't just an exercise in programming AI, but an analysis of human behavior.
 
You must log in or register to reply here.
Back
Top