N E U R A L
N E T W O R K
S T Y L E
T R A N S F E R S
I’ve heard people both in business and in academia say that artificial neural networks are useless, which I believe is an astounding claim given how much practical application we've successfully found for them so early in the technology's development. I think this misguided sentiment stems from not understanding how to qualify whether deep learning is appropriate for a given application or data set. To better understand how common issues like overtraining and noise affect feature recognition, I’ve been toying around with TensorFlow, Google's end-to-end open source machine learning platform.
One of deep learning's more whimsical applications is to conduct a style transfer, wherein a neural network trained to recognize features in one painting can attempt to interpret those same features in another image, thereby repainting it in the style of the training image.
These are some of my more successful results from these experiments.
Below is the original photograph I captured of Annex House's door as reference.
W O R L D T R A D E
C E N T E R
This tribute was created on September 11, 2020 by feeding a collage I had created from images of textures into an artificial neural network.
B O D E G A W I N D O W
D E R E L I C T L O T
This is the original photograph.
