CocoDetection() and CocoCaptions() should be replaced with MSCoco()

[hyperkai]

🚀 The feature

It seems like CocoDetection() and CocoCaptions() are identical(same) according to the experiments which I did and the source code. *CocoDetection() is just renamed to CocoCaptions() by extending CocoDetection().

Motivation, pitch

So, CocoDetection() and CocoCaptions() should be replaced with MSCoco().

Alternatives

No response

Additional context

No response

[NicolasHug]

Hi @hyperkai , can you share a minimal reproducing example showing that the output is exactly the same? The targets of CocoCations should be different:

vision/torchvision/datasets/coco.py

Line 108 in 11fd635

def _load_target(self, id: int) -> List[str]: