- class mmpretrain.evaluation.VOCMultiLabelMetric(*arg, difficult_as_positive=None, **kwarg)¶
A collection of metrics for multi-label multi-class classification task based on confusion matrix for VOC dataset.
It includes precision, recall, f1-score and support.
difficult_as_postive (Optional[bool]) – Whether to map the difficult labels as positive in one-hot ground truth for evaluation. If it set to True, map difficult gt labels to positive ones(1), If it set to False, map difficult gt labels to negative ones(0). Defaults to None, the difficult labels will be set to ‘-1’.
**kwarg – Refers to MultiLabelMetric for detailed docstrings.