In metric_callbacks.py, lines 34-36
def _metrics_factory(self, by_class=False): average = None if by_class else "micro" average_iou = None if by_class else "macro" # special case, only mean IoU is of interest

This results in equal metrics of overall accuracy, recall, precision and f1-score. Wouldn't it be better to average 'macro', as the resulting metrics would be unambiguous?
(except for accuracy)