Skip to content

Latest commit

 

History

History
28 lines (16 loc) · 1.11 KB

README.md

File metadata and controls

28 lines (16 loc) · 1.11 KB

Art-Generation

This project was built to generate 1 single piece of art(picture) by merging components from 2 input images. It makes use of Neural Style Transfer.

Neural Style Transfer (NST) uses a previously trained convolutional network, and builds on top of that. The idea of using a network trained on a different task and applying it to a new task is called transfer learning.

NST uses a previously trained convolutional network, and builds on top of that. Transfer learning is an idea, in which we use a network trained on a different task and apply it to a new task.

Computing the Cost function for content image Jcontent(C,G)

Gram Matrix: Also called the 'style matrix'. It compares how similar two vectors are, using np.dot()

Content Image Used by me:

Style Image Used by me:

Generated Image: