add HMDB51 and UCF101 datasets as well as prototype for new style video decoding

[pmeier]

This adds HMDB51 as first video dataset (see #4541). It is used for prototyping video decoder datapipes initially implemented by @bjuncek in #4838. You can test with the following snippet

from torchvision.prototype import datasets

dataset = datasets.load("hmdb51")
dataset = datasets.utils.KeyframeDecoder(dataset)

for sample in dataset:
    print(sample)
    break

Apart from that the biggest change is the added meta dictionary attribute to the EncodedData feature. The rationale here is that each dataset might provide very different meta data for each file and this is hard to standardize. We might be able to have some common attributes like path, but I would still leave the option open for arbitrary meta data.

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 07d78b2 (more details on the Dr. CI page):

---

• 1/1 failures introduced in this PR

---

🕵️ 1 new failure recognized by patterns

The following CI failures do not appear to be due to upstream breakages:

unittest_prototype (1/1)

Step: "Run tests" (full log | diagnosis details | 🔁 rerun)

�[31m============================== �[31m�[1m5 ...0.26s�[0m�[31m ===============================�[0m

    raise TypeError(f"{cls} is not a generic class")
TypeError: <class 'torch.utils.data.datapipes.iter.grouping.ShardingFilterIterDataPipe'> is not a generic class
------ generated xml file: /home/circleci/project/test-results/junit.xml -------
=========================== short test summary info ============================
ERROR test/test_prototype_builtin_datasets.py - TypeError: <class 'torch.util...
ERROR test/test_prototype_datasets_api.py - TypeError: <class 'torch.utils.da...
ERROR test/test_prototype_datasets_utils.py - TypeError: <class 'torch.utils....
ERROR test/test_prototype_models.py - TypeError: <class 'torch.utils.data.dat...
ERROR test/test_prototype_transforms_functional.py - TypeError: <class 'torch...
!!!!!!!!!!!!!!!!!!! Interrupted: 5 errors during collection !!!!!!!!!!!!!!!!!!!!
�[31m============================== �[31m�[1m5 errors�[0m�[31m in 0.26s�[0m�[31m ===============================�[0m


Exited with code exit status 2

---

This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

[pmeier]

@bjuncek I've also ported UCF101 from #4848 so you can close it and take over this PR.

[pmeier]

@NicolasHug The archive is a rar of rars so using a single extract=True won't cut it. We need the full extraction since reading from rar archives is rather slow and with this we get a significant performance increase.

Another option would be to use this "recursive extraction" by default when setting extract=True.

[bjuncek]

@pmeier I see that you closed this; should I switch to the new PR?

[pmeier]

This probably happened because I deleted the branch that this was supposed to be merged into. I'll fix, keep working on this.

[pmeier]

For whatever reason, GitHub does not let me re-open this. I've send the PR again in #5422.

[datumbox]

@pmeier It's because the revamp branch is deleted.

[pmeier]

Yeah, I figured that, but it doesn't let me select a new merge target like main either. This is possible for open PRs.