Redundant return in `convert_format`

[ariG23498]

In the following lines I have noticed that the bounding boxes are converted in place. I am not sure why we return the dictionary again. This does not raise a bug, but breaks the mental model of inplace and out of place changes.

keras-cv/keras_cv/bounding_box/converters.py

Lines 372 to 381 in 925d99d

	if isinstance(boxes, dict):
	boxes["boxes"] = convert_format(
	boxes["boxes"],
	source=source,
	target=target,
	images=images,
	image_shape=image_shape,
	dtype=dtype,
	)
	return boxes

It would rather be useful to create a new dictionary and return the converted bbox where the original is kept as is.

If that is a good change, I would like to contribute to this (will be a quick fix)

[LukeWood]

Agreed - seems better to abide by functional programming paradigms in order to match tf/jax/torch's API styles!

I thought this copied the boxes, but can't find a reference to boxes.copy() in the git history!

[divyashreepathihalli]

@ariG23498 Thanks for filing the issue! please go ahead and open a PR!