rrmina / fast-neural-style-pytorch

Fast Neural Style Transfer implementation in PyTorch :art: :art: :art:
317 stars 77 forks source link

Train Notebook: Request for Improvements #3

Closed rrmina closed 5 years ago

rrmina commented 5 years ago

This still needs a lot of changes to make it a one-click-run-all interface. Some of the changes requested are:

This notebook can be used a reference for the first 4 bullets. Can you do it @p-rit ?

p-rit commented 5 years ago

Will try to improve

On Fri, 18 Jan 2019, 5:05 am Rusty Mina <notifications@github.com wrote:

This still needs a lot of changes to make it a one-click-run-all interface. Some of the changes requested are:

  • First cell should contain all of the wheel installations and
  • First cell should also download all of the required local files (e.g. image file)
  • Images should be displayed every show_iter
  • A link directing to Google Colab is also welcome.

This notebook https://github.com/iamRusty/neural-style-pytorch/blob/master/neural_style.ipynb can be used a reference for the first 3 bullets. Can you do it @p-rit https://github.com/p-rit ?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/iamRusty/fast-neural-style-pytorch/issues/3, or mute the thread https://github.com/notifications/unsubscribe-auth/Ai_AbEJCVJVJ73_cvNhOMA1qQZ1Pd9vjks5vEQi-gaJpZM4aG8l_ .

p-rit commented 5 years ago

link directing to google colab notebook containing improved code

https://colab.research.google.com/drive/1lL81ANjjehOAudXy6CHNUHmQ9aFdJvLg#scrollTo=KhEWjKxMb0oO

rrmina commented 5 years ago

Hi @p-rit . Thank you improving the notebook. Some comments:

!wget https://github.com/iamRusty/fast-neural-style-pytorch/blob/master/transformer.py

do

!wget https://raw.githubusercontent.com/iamRusty/fast-neural-style-pytorch/master/transformer.py

* You don't need to put the cell containing itot, ttoi, etc. `utils.py` already has the more updated version of the helper codes. So yeah, you may **delete** that cell.

* Lastly, display the image after saving the sample image. You may use:

utils.show(sample_image)



Edit: One more thing, there's a typo with the import of the `models` from `torchvision`
Edit2: **Also, change the content weight and style weight to `2e-6` and `50` respectively. The current settings were used for udnie, not mosaic. the mosaic style needs more weight on content :)**