Codecov Report

Merging #851 (d30902a) into main (096eb48) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #851   +/-   ##
=======================================
  Coverage   94.84%   94.84%           
=======================================
  Files         133      133           
  Lines        5200     5200           
=======================================
  Hits         4932     4932           
  Misses        268      268           

Flags with carried forward coverage won't be shown. Click here to find out more.

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 096eb48...d30902a. Read the comment docs.

That's a very interesting PR :)

I have to admit, it would have been nice to split it in a few haha

General comments first:

• the code of conduct & contributing markdown cannot be moved (it's for github). But if you need them someplace else, we can create a symlink
• could you explain the new dependencies of the documentation please?
• for the dataset images, I'll add them in a release attachments to limit the amount of content files in the repo
• I see that you use tabs for PyTorch/TensorFlow: it's really neat, I'd love to have that. However, I didn't see how you mean to build this for both framework (you can't build for both of them at once)

@fg-mindee
Yes i know but the most stuff in this PR depends on other so really hard to minimize 😅

1. I agree but i would say we hold this also in the documentation so we have this twice but can link in the readme directly to docs. Wdyt?
2. furo is a theme like the old sphinx_rtd_theme and sphinx-tabs is an sphinx extension which i have used to realize the tf/pytorch Code example split :)
3. 👍 than we need only change to the url
4. Yes i know i have take care of this directly in the docstrings you can check this in any changed model file (for example resnet) :D

@fg-mindee
Point 1: applied as symlinks and reverted both files 👍

Point 3: zip file with dataset image-grids attatched
dataset_image_grid_examples.zip

Point 4:
Every object which shares the same namespace in TF and PT implementation (and is also implemented same : which makes the transform part a bit more tricky) has now a shared docstring (same in TF and PT implementation).
Hopefully this makes a bit more clear how it is possible that we have this tab seperated code examples 🤗

I have added es for both TF implementation and PT. Why? :

1. if we want we could easy switch "as backend" to pt for the docs without any loss
2. consistency !
   I have tested this docstrings also while coding and its really nice to have both samples in one way i like it 😅
   Disadvantage are a bit longer docstrings

.. tabs::

        .. tab:: TensorFlow

            .. code:: python

                >>> import tensorflow as tf
                >>> from doctr.models import vgg16_bn_r
                >>> model = vgg16_bn_r(pretrained=False)
                >>> input_tensor = tf.random.uniform(shape=[1, 512, 512, 3], maxval=1, dtype=tf.float32)
                >>> out = model(input_tensor)

        .. tab:: PyTorch

            .. code:: python

                >>> import torch
                >>> from doctr.models import vgg16_bn_r
                >>> model = vgg16_bn_r(pretrained=False)
                >>> input_tensor = torch.rand((1, 3, 512, 512), dtype=torch.float32)
                >>> out = model(input_tensor)

Would it be possible to split this into several PRs please? 🙏

• one for the change of theme + JS/CSS
• updating the docstrings of datasets with images
• the rest in a final PR

That would help reducing the amount of files modified here 😅

Now that all the subPRs have been merged, this PR can be closed :)
Thanks a lot for your time @felixdittrich92 🙏 This is a huge improvement to the general look of the documentation!!

@felixdittrich92: thanks for this update, I really love the new visual!