Skip to content

Latest commit

 

History

History
8 lines (8 loc) · 710 Bytes

README.md

File metadata and controls

8 lines (8 loc) · 710 Bytes

Neural-Style-Transfer

  • This repository is a style transfer using a vgg19 classifier which is a premade neural network in pytorch with 36 layers
  • The series of layers of convolution and pooling layers are made along with drouput regularization
  • using the gram loss , style of the second image is copied in due course of the epoch of neural net
  • using the MSE loss function , the difference between target image and the initiaal image is calculated
  • then there are weights assigned so that there is a resolution to the problem between the tradeoff for style transfer and keeping . the identity of the image.
  • reference : https://github.com/uda.city/deep-learning-v2-pytorch/tree/master/style-transfer