# assigners

Python API: `luxonis_train.assigners`

Assigners that match the anchors of a detection head to the ground truth boxes.

An assigner decides which anchors predict each ground truth box. It then builds the per-anchor labels, boxes, and scores that the
detection losses read. The package has two assigners:

 * [ATSSAssigner](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/assigners/atss_assigner.md)
   selects the positive anchors from the anchor geometry and an adaptive IoU threshold.
 * [TaskAlignedAssigner](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/assigners/tal_assigner.md)
   selects the positive anchors by a metric that combines the predicted class score and the IoU of the predicted box.

## Child Pages

 * [atss_assigner](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/assigners/atss_assigner.md)
 * [tal_assigner](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/assigners/tal_assigner.md)
 * [utils](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/assigners/utils.md)