Metrics

Segmentation metrics

compute_meandice

compute_meandice.compute_meandice(y, include_background=True, to_onehot_y=True, mutually_exclusive=False, add_sigmoid=False, logit_thresh=0.5)

Computes dice score metric from full size Tensor and collects average.

Parameters

  • y_pred (torch.Tensor) – input data to compute, typical segmentation model output. it must be One-Hot format and first dim is batch, example shape: [16, 3, 32, 32].

  • y (torch.Tensor) – ground truth to compute mean dice metric, the first dim is batch. example shape: [16, 1, 32, 32] will be converted into [16, 3, 32, 32]. alternative shape: [16, 3, 32, 32] and set to_onehot_y=False to use 3-class labels directly.

  • include_background (Bool) – whether to skip Dice computation on the first channel of the predicted output. Defaults to True.

  • to_onehot_y (Bool) – whether to convert y into the one-hot format. Defaults to True.

  • mutually_exclusive (Bool) – if True, y_pred will be converted into a binary matrix using a combination of argmax and to_onehot. Defaults to False.

  • add_sigmoid (Bool) – whether to add sigmoid function to y_pred before computation. Defaults to False.

  • logit_thresh (Float) – the threshold value used to convert (after sigmoid if add_sigmoid=True) y_pred into a binary matrix. Defaults to 0.5.

Returns

[batch_size, n_classes]).

Return type

Dice scores per batch and per class (shape

Note

This method provides two options to convert y_pred into a binary matrix

  1. when mutually_exclusive is True, it uses a combination of argmax and to_onehot,

  2. when mutually_exclusive is False, it uses a threshold logit_thresh (optionally with a sigmoid function before thresholding).