FastSurferCNN.utils.meters¶

class FastSurferCNN.utils.meters.Meter(cfg, mode, global_step, total_iter=None, total_epoch=None, class_names=None, device=None, writer=None)[source]¶

Meter class to keep track of the losses and scores during training and validation.

Methods

`disable_confusion_mat`()	Disable confusion matrix.
`enable_confusion_mat`()	Enable confusion matrix.
`log_epoch`(cur_epoch)	Log the current epoch.
`log_iter`(cur_iter, cur_epoch)	Log the current iteration.
`reset`()	Reset bach losses and dice scores.
`update_stats`(pred, labels, batch_loss)	Update the statistics.
`write_summary`(loss_total[, lr, loss_ce, ...])	Write a summary of the losses and scores.

disable_confusion_mat()[source]¶: Disable confusion matrix.

enable_confusion_mat()[source]¶: Enable confusion matrix.

log_epoch(cur_epoch)[source]¶

Log the current epoch.

Parameters:

cur_epochint: Current epoch.

log_iter(cur_iter, cur_epoch)[source]¶

Log the current iteration.

Parameters:

cur_iterint: Current iteration.
cur_epochint: Current epoch.

reset()[source]¶: Reset bach losses and dice scores.

update_stats(pred, labels, batch_loss)[source]¶: Update the statistics.

write_summary(loss_total, lr=None, loss_ce=None, loss_dice=None)[source]¶

Write a summary of the losses and scores.

Parameters:

loss_totaltorch.Tensor: Total loss.
lrdefault = None: Learning rate (Default value = None).
loss_cedefault = None: Cross entropy loss (Default value = None).
loss_dicedefault = None: Dice loss (Default value = None).