MeanAveragePrecision does not handle empty predictions or ground truth boxes not in xyxy format

[kouyk]

🐛 Bug

When empty ground truth or prediction boxes tensor are provided, MeanAveragePrecision crashes during the conversion to xyxy from other formats.

To Reproduce

Steps to reproduce the behavior...

1. Have empty boxes tensor in ground truth or prediction
2. instantiate MeanAveragePrecision with box_format='xywh'
3. run update with the ground truth and prediction boxes
4. conversion error

Code sample

import torch
from torchmetrics.detection.mean_ap import MeanAveragePrecision
gts = [
        dict(boxes=torch.Tensor([[10, 20, 15, 25]]), labels=torch.IntTensor([0])),
    ]
preds = [
    dict(boxes=torch.Tensor([]), scores=torch.Tensor([]), labels=torch.IntTensor([])),
]
metric = MeanAveragePrecision(box_format='xywh')
metric.update(preds, gts)

Expected behavior

torchmetrics should skip the conversion for empty boxes.

Environment

• TorchMetrics version (and how you installed TM, e.g. conda, pip, build from source): 0.9.1 from nix
• Python & PyTorch Version (e.g., 1.0): Python 3.9, PyTorch 1.11.0
• Any other relevant information such as OS (e.g., Linux): Linux

Additional context

[github-actions]

Hi! thanks for your contribution!, great first issue!