[FR] Allow mask_path in Detection labels

[Laurent2916]

Proposal Summary

Currently Detection only accepts mask fields, which are serialized numpy arrays saved to the mongo database. It would instead be useful to be able to instruct fiftyone to use masks saved on disk, with a field mask_path, just like how Segmentation does:

fiftyone/fiftyone/core/labels.py

Lines 1006 to 1021 in 39201f4

	class Segmentation(_HasID, _HasMedia, Label):
	"""A semantic segmentation for an image.
	Provide either the ``mask`` or ``mask_path`` argument to define the
	segmentation.
	Args:
	mask (None): a numpy array with integer values encoding the semantic
	labels
	mask_path (None): the absolute path to the segmentation image on disk
	"""
	_MEDIA_FIELD = "mask_path"
	mask = fof.ArrayField()
	mask_path = fof.StringField()

Motivation

This would make it possible to offload some disk usage from the mongo database to somewhere else.

What areas of FiftyOne does this feature affect?

• App: FiftyOne application
• Core: Core fiftyone Python library
• Server: FiftyOne server

Details

I'm not quite familiar with the codebase yet, though from looking at Segmentation it doesn't look that hard. I would appreciate a maintainer's opinion on the feasibility of this feature.

Willingness to contribute

• Yes. I can contribute this feature independently
• Yes. I would be willing to contribute this feature with guidance from the FiftyOne community
• No. I cannot contribute this feature at this time

[Laurent2916]

Closed by #5120