PyTorch implementation of A Neural Algorithm of Artistic Style by Leon A. Gatys, et al.
This is the progress of the first few iterations
- The hyperparameters are same as used in the paper.
- All the code was written and ran on Google Colab.
- As in the paper, conv1_1, conv2_1, conv3_1, conv4_1, conv5_1 are used for style loss.
- SImilarly, conv4_2 is used for content loss.
- PyTorch's forward hooks are used to extract intermediate feature maps.
- VGG19's MaxPool2d layers were replaced by AvgPool2d layers.
torch
torchvision
numpy
matplotlib
Pillow
imageio
- Leon A. Gatys, et al. A Neural Algorithm of Artistic Style [arxiv]