Add Mask R-CNN

[NielsRogge]

What does this PR do?

This PR adds the classic Mask R-CNN framework for object detection and instance segmentation.

To do/to be discussed:

• [ ] where to place utilities like NMS, loss computation, samplers
• [ ] whether to create dummies for torchvision-backed models
• [ ] how to add support for the object detection pipeline - either add **kwargs to each post_process_object_detection method, or add specific logic for Mask R-CNN inside object_detection_pipeline.py

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint.

[amyeroberts]

@NielsRogge As @sgugger mentions, the PR is still in WIP state. Happy to review once transformers ready :)

[NielsRogge]

I've updated all docstrings and variable names, PR is ready for another review

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[aymanechilah]

Hello @NielsRogge, what is the status of this feature? Thanks in advance

[bit-scientist]

Hi, @NielsRogge looking forward to it. Could you, for now, recommend a robust text detector available here to combine with TrOCR. I would like to see how well the two work with the help of HF🤗.