transfer learning

Reusing a model's weights, pretrained on one task, as the starting point for a new task by retraining only part of it.

Transfer learning reuses a backbone's weights, pretrained on a large source task, as the starting point for a new target task -- typically by freezing the pretrained layers and training only a new task-specific head on top -- letting a task with limited labeled data still benefit from features learned on a much larger dataset elsewhere.

Examples

A backbone pretrained on a million general photos is frozen, and only a brand-new 5-class head is trained on a small custom flower dataset.