woldier

https://github.com/open-mmlab/mmsegmentation/blob/b040e147adfa027bbc071b624bedf0ae84dfc922/configs/_base_/datasets/loveda.py#L10C2-L10C27 The original size of the image in LoveDa is 1024*1024 why the img_scale is set to (2048, 512) here?

https://github.com/lhoyer/MIC/blob/2f932a98b5dd9f598aaeb32411863ceea0809314/seg/mmseg/apis/train.py#L71 In case of parallel training, then DCAS.clas will be wrapped by `DistributedDataParallelWrapper`, where the model will be wrapped by `MMDistributedDataParallel`, by reading the `MMDistributedDataParallel` source code, it can be...

Segmentation IOU compute Ignore some tagged values that don't need to be recorded (such as 255)

1

## 🚀 Feature when we compute IOU ```python import torch _ = torch.manual_seed(0) from torchmetrics.segmentation import MeanIoU miou = MeanIoU(num_classes=3) preds = torch.randint(0, 2, (5,)) target = torch.as_tensor((0, 1, 2,...