Losses registry

Loss functions assembled by build_criteria(). See Add a model or loss.

Build any of these from a config with criteria=[dict(type=...)], using the type string in the first column. Generated from the live registry - 14 classes (14 names including aliases), grouped by role.

Classification

Cross-entropy and focal variants for class logits.

Name	Summary
BinaryFocalLoss	Binary focal loss for class-imbalanced binary targets.
CrossEntropyHeadLoss	Cross-entropy for an extra categorical per-query head.
CrossEntropyLoss	Standard multi-class cross-entropy over per-point class logits.
FocalLoss	Multi-class focal loss over per-point class logits.
SmoothCELoss	Label-smoothed cross-entropy with NaN-robust averaging.

pimm.models.losses.misc.BinaryFocalLoss	Binary focal loss for class-imbalanced binary targets.
pimm.models.losses.misc.CrossEntropyHeadLoss	Cross-entropy for an extra categorical per-query head.
pimm.models.losses.misc.CrossEntropyLoss	Standard multi-class cross-entropy over per-point class logits.
pimm.models.losses.misc.FocalLoss	Multi-class focal loss over per-point class logits.
pimm.models.losses.misc.SmoothCELoss	Label-smoothed cross-entropy with NaN-robust averaging.

Segmentation

Region-overlap losses for semantic segmentation.

Name	Summary
DiceLoss	Soft multi-class Dice loss over per-point class probabilities.
LovaszLoss	Lovasz loss: a direct surrogate for the IoU (Jaccard) metric.

pimm.models.losses.misc.DiceLoss	Soft multi-class Dice loss over per-point class probabilities.
pimm.models.losses.lovasz.LovaszLoss	Lovasz loss: a direct surrogate for the IoU (Jaccard) metric.

Regression

Continuous-target losses (momentum, vertex, …).

Name	Summary
L1RegressionLoss	Mean-absolute-error (L1) regression loss for continuous targets.
MSERegressionLoss	Mean-squared-error (L2) regression loss for continuous targets.
SmoothL1RegressionLoss	Smooth-L1 (Huber) regression loss for continuous targets.

pimm.models.losses.misc.L1RegressionLoss	Mean-absolute-error (L1) regression loss for continuous targets.
pimm.models.losses.misc.MSERegressionLoss	Mean-squared-error (L2) regression loss for continuous targets.
pimm.models.losses.misc.SmoothL1RegressionLoss	Smooth-L1 (Huber) regression loss for continuous targets.

Instance segmentation

Set-based mask + class (+ regression) losses with Hungarian matching.

Name	Summary
FastInstanceSegmentationLoss	Mask2Former-style instance segmentation loss (cached, vectorized).
FastInstanceSegmentationRegressionLoss	Instance segmentation loss plus configurable per-instance regressions.
FastUnifiedInstanceLoss	Combine instance mask/class losses with configured query-head losses.
InstanceSegmentationLoss	DEPRECATED alias for FastInstanceSegmentationLoss.

pimm.models.losses.instance_fast.FastInstanceSegmentationLoss	Mask2Former-style instance segmentation loss (cached, vectorized).
pimm.models.losses.instance_regression_fast.FastInstanceSegmentationRegressionLoss	Instance segmentation loss plus configurable per-instance regressions.
pimm.models.losses.instance_unified_fast.FastUnifiedInstanceLoss	Combine instance mask/class losses with configured query-head losses.
pimm.models.losses.instance_fast.InstanceSegmentationLoss	DEPRECATED alias for FastInstanceSegmentationLoss.