[videoAPI] minor example and documentation changes

[bjuncek]

There were still a few points where documentation should have been clarified, and
example notebook needed to change the return type. Should be cleared up now.

[vfdev-5]

@bjuncek on Colab running the notebook on ('1.8.0.dev20201015+cpu', '0.8.0.dev20201015+cpu') from pypi, I have the following error:

/usr/local/lib/python3.6/dist-packages/torchvision/io/__init__.py in __init__(self, path, stream)
     88     def __init__(self, path, stream="video"):
     89         if not _has_video_opt():
---> 90             raise RuntimeError("Not compiled with video_reader support")
     91         self._c = torch.classes.torchvision.Video(path, stream)
     92 

RuntimeError: Not compiled with video_reader support

What do you think to add a message about to install manually ffmpeg and build torchvision from tag to have the feature ?
Idea is to guide the user a bit more...

• Another point, probably, a more generic question. In Colab you are make_dataset method which is not documented at all.
  Maybe, it worth adding a docstring.

• it would be nice to add an image of frames from the batch in the end of the notebook.

import matplotlib.pylab as plt
%matplotlib inline

plt.figure(figsize=(12, 12))
for i in range(16):
  plt.subplot(4, 4, i + 1)
  plt.imshow(b["video"][11, i, ...].permute(1, 2, 0))
  plt.axis("off")

[bjuncek]

Hi @vfdev-5

What do you think to add a message about to install manually ffmpeg and build torchvision from tag to have the feature?

Since there is a warning there already, might as well. Added in subsequent commit.

Another point, probably, a more generic question. In Colab you are make_dataset method which is not documented at all.

Isn't it documented in the DatasetFolder?

it would be nice to add an image of frames from the batch in the end of the notebook.

done :)

[vfdev-5]

LGTM ! Thanks @bjuncek

Failing Travis CI is unrelated.

[codecov]

Codecov Report

Merging #2821 into master will decrease coverage by 1.08%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master    #2821      +/-   ##
==========================================
- Coverage   73.35%   72.27%   -1.09%     
==========================================
  Files          99       99              
  Lines        8785     8785              
  Branches     1387     1387              
==========================================
- Hits         6444     6349      -95     
- Misses       1916     1998      +82     
- Partials      425      438      +13     

Impacted Files	Coverage Δ
torchvision/io/__init__.py	89.65% <0.00%> (ø)
torchvision/ops/_register_onnx_ops.py	48.78% <0.00%> (-36.59%)	⬇️
torchvision/models/detection/transform.py	78.33% <0.00%> (-17.23%)	⬇️
torchvision/ops/poolers.py	85.29% <0.00%> (-11.77%)	⬇️
torchvision/ops/boxes.py	87.35% <0.00%> (-8.05%)	⬇️
torchvision/models/detection/roi_heads.py	77.23% <0.00%> (-5.11%)	⬇️
torchvision/models/detection/rpn.py	90.18% <0.00%> (-3.69%)	⬇️

---

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 99bf4c0...8396963. Read the comment docs.

[fmassa]

Thanks for the PR!

I have one blocking comment (the doc is now broken), otherwise the other comments are minor.

[fmassa]

Thanks!