SetPredictionClassificationLoss

Enums · AiDotNet.Enums

The classification objective used by a DETR-family set prediction loss.

For Beginners

A DETR-style detector predicts a fixed set of candidate objects. After each ground-truth object is matched to one candidate, this setting chooses how the class scores of every candidate are trained.

Fields

Field Summary
SetPredictionClassificationLoss SigmoidFocal Per-class sigmoid focal loss with no no-object class (Lin et al.
SetPredictionClassificationLoss SoftmaxCrossEntropy Softmax cross-entropy over the foreground classes plus a final no-object class, with the no-object class down-weighted (Carion et al.
SetPredictionClassificationLoss VariFocal IoU-aware varifocal loss (Zhang et al.
int value__