Transfer learning with VGG19 using torchvision

[sitamgithub-MSIT]

This notebook applies transfer learning with the "ConvNet as fixed feature extractor" approach. vgg19 is the model used for the base-pretrained model.

Additional Changes Made

• Added remarks to several different code sections.
• Rather than placing everything in one function, distinct helper functions were implemented for training.
• Improved interpretibility of functions by assigning corresponding document strings. Additionally, data augmentation and processing were addressed.

cc: @merveenoyan, @MKhalusova

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[sitamgithub-MSIT]

I faced a problem while using Plotly in which the image needed to be static to display inside the notebook. To resolve this, I used a picture of my graph and created a pull request for it on huggingface-cv-org. I attempted to use other contributor images just for testing, but they were not rendering properly inside the notebook. As a solution, I placed the picture inside my working unit under the notebook folder.

[sitamgithub-MSIT]

I faced a problem while using Plotly in which the image needed to be static to display inside the notebook. To resolve this, I used a picture of my graph and created a pull request for it on huggingface-cv-org. I attempted to use other contributor images just for testing, but they were not rendering properly inside the notebook. As a solution, I placed the picture inside my working unit under the notebook folder.

Though in ReviewNB, the plotly graph is visible.

[MKhalusova]

Thank you for working on this! Loved the style, and detailed explanations. I think we can skip the distribution plot in this example. Left a couple of comments in the notebook too.

[sitamgithub-MSIT]

thanks a lot for working on this 💜

As both the torchvision part are almost ready and for vgg 19 reviewer comments are also addressed almost. So you can expect vgg19 to complete by eod. And for resnet also as the overall structure is same so shouldn't be any problem there as well.

[sitamgithub-MSIT]

@merveenoyan, except for two visualization problems, one is in plot distribution and another is in prediction. Everything I pushed into the latest commit. For those two viz changes, I need to modify the code and run the entire notebook again. If all the content is right, then I can make another final branch for vgg19 with those viz changes just for merge purposes, as most of the changes have already been made.

[merveenoyan]

it seems ready to merge to me, if @MKhalusova has no comments :)

[sitamgithub-MSIT]

@merveenoyan It is ready to be merged!

[johko]

Merging :rocket: