open-mmlab/mmdetection · error · KeyError

metrics {iou_metrics} is not supported. Only supports mIoU/m

Error message

metrics {iou_metrics} is not supported. Only supports mIoU/mDice/mFscore.

What it means

IoUMetric (semantic segmentation) validates iou_metrics is a subset of {'mIoU','mDice','mFscore'}; anything else raises this KeyError at construction. The argument may be a single string or list, but only those three statistic names are implemented.

Source

Thrown at mmdet/evaluation/metrics/semseg_metric.py:62

            names to disambiguate homonymous metrics of different evaluators.
            If prefix is not provided in the argument, self.default_prefix
            will be used instead. Defaults to None.
    """

    def __init__(self,
                 iou_metrics: Sequence[str] = ['mIoU'],
                 beta: int = 1,
                 collect_device: str = 'cpu',
                 output_dir: Optional[str] = None,
                 format_only: bool = False,
                 backend_args: dict = None,
                 prefix: Optional[str] = None) -> None:
        super().__init__(collect_device=collect_device, prefix=prefix)

        if isinstance(iou_metrics, str):
            iou_metrics = [iou_metrics]
        if not set(iou_metrics).issubset(set(['mIoU', 'mDice', 'mFscore'])):
            raise KeyError(f'metrics {iou_metrics} is not supported. '
                           f'Only supports mIoU/mDice/mFscore.')
        self.metrics = iou_metrics
        self.beta = beta
        self.output_dir = output_dir
        if self.output_dir and is_main_process():
            mkdir_or_exist(self.output_dir)
        self.format_only = format_only
        self.backend_args = backend_args

    def process(self, data_batch: dict, data_samples: Sequence[dict]) -> None:
        """Process one batch of data and data_samples.

        The processed results should be stored in ``self.results``, which will
        be used to compute the metrics when all batches have been processed.

        Args:
            data_batch (dict): A batch of data from the dataloader.
            data_samples (Sequence[dict]): A batch of outputs from the model.

View on GitHub (pinned to cfd5d3a985)

Solutions

1. Restrict iou_metrics to mIoU, mDice, mFscore (any subset, exact casing)
2. Remove mAcc/aAcc from the list; note overall accuracy is reported separately by the metric where supported
3. Set iou_metrics='mIoU' (the common default) if unsure

Example fix

# before
val_evaluator = dict(type='IoUMetric', iou_metrics=['mIoU', 'mAcc', 'aAcc'])
# after
val_evaluator = dict(type='IoUMetric', iou_metrics=['mIoU'])

SUPPORTED = {'mIoU','mDice','mFscore'}
bad = set(iou_metrics if isinstance(iou_metrics, list) else [iou_metrics]) - SUPPORTED
assert not bad, f'unsupported iou_metrics: {bad}'

def are_valid_iou_metrics(v) -> bool:
    items = [v] if isinstance(v, str) else v
    return isinstance(items, list) and set(items).issubset({'mIoU','mDice','mFscore'})

When it happens

Trigger: Passing iou_metrics=['mIoU','mAcc','aAcc'] or iou_metrics='dice' (wrong casing/name) to IoUMetric; confusing semanticseg metrics with classification metrics.

Common situations: Copy-pasting metric lists from mmcls or older mmseg configs (mAcc/aAcc moved elsewhere or removed); misspelling 'mFscore' as 'f1' or 'mF1'.

Related errors

• metric must be a list or a str.
• out_indices must be a subset of range(0, 8). But received {o
• Expect "arch" to be either a string or a dict, got {type(arc
• num_classes={num_classes} is too small
• Invalid text mode "{self.text_mode}".

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/352aa48d47e06c99. Report an issue: GitHub.