The stock photography search engine Everypixel has launched a new web app called Everypixel Aesthetics that uses a neural network to tell you how awesome any photo is.
“We trained neural networks to see the beauty of photos the same way as you and I do,” Everypixel tells PetaPixel. “To handle it, designers, editors and experienced stock photographers helped us generate a training dataset with 956,794 positive and negative patterns.”
Using the dataset, the service then takes a look at any photo you provide and returns a percent chance indicating the likelihood that the photo is “awesome.” Alongside the percentage, the site also automatically generates a list of tags that may be relevant to what’s seen in the shot.
While the idea is certainly interesting, the some of the results we got were a bit puzzling.
It seems the neural network didn’t think very highly of Ansel Adams’ classic “The Tetons and the Snake River (1942),” giving it only a 32.3% chance of being an awesome shot:
By comparison, the system judged that a random cat photo was much more awesome:
It did consider a stock photo of bell peppers to be not so great:
Strangely enough, the highest rated shot we submitted turned out to be a very random photo of a bird sitting on a swimming pig. This photo apparently has a 99.9% chance of being awesome:
So if you’ve been looking to create a photo series that launches your photography career into the stratosphere, see if you can get your hands on a bird and a pig.
from PetaPixel https://petapixel.com/2017/04/08/everypixel-aesthetics-uses-rate-photos-awesomeness/
No comments:
Post a Comment