# heads

Python API: `luxonis_train.nodes.heads`

将特征转换为预测的头部。

每个头部在其 `task` 类属性中设置一个任务。任务决定了哪些损失、指标和可视化工具可以附加到头部。`parser` 类属性命名了解析器，该解析器读取导出模型中头部的输出。没有导出解析器的头部保留空的默认 `""`。

按任务分组的头部：

 * 分类：[ClassificationHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/classification_head.md)
   和
   [TransformerClassificationHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/transformer_classification_head.md)。
 * 分割：[SegmentationHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/segmentation_head.md)、[BiSeNetHead](https://docs.luxonis.com/zh/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/bisenet_head.md)、[DDRNetSegmentationHead](https://docs.luxonis.com/zh/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/ddrnet_segmentation_head.md)
   和
   [TransformerSegmentationHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/transformer_segmentation_head.md)。
 * 边界框：[EfficientBBoxHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/efficient_bbox_head.md)
   和
   [PrecisionBBoxHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/precision_bbox_head.md)。
 * 实例关键点：[EfficientKeypointBBoxHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/efficient_keypoint_bbox_head.md)。
 * 实例分割：[PrecisionSegmentBBoxHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/precision_seg_bbox_head.md)。
 * FOMO
   对象中心：[FOMOHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/fomo_head.md)。
 * 异常检测：[DiscSubNetHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/discsubnet_head/discsubnet_head.md)。
 * OCR：[OCRCTCHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/ocr_ctc_head.md)。
 * 嵌入：[GhostFaceNetHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/ghostfacenet_head.md)。

边界框、实例关键点和实例分割头部派生自
[BaseDetectionHead](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/base_detection_head.md)。它们仅在评估模式下运行非极大值抑制
(NMS)。训练模式和导出模式跳过 NMS。指标在评估模式下运行。因此 `conf_thres`、`iou_thres` 和 `max_det` 会改变指标所看到的内容。NN Archive 为导出解析器存储相同的三个值。

## 子页面

 * [discsubnet_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/discsubnet_head.md)
 * [base_detection_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/base_detection_head.md)
 * [base_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/base_head.md)
 * [bisenet_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/bisenet_head.md)
 * [classification_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/classification_head.md)
 * [ddrnet_segmentation_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/ddrnet_segmentation_head.md)
 * [efficient_bbox_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/efficient_bbox_head.md)
 * [efficient_keypoint_bbox_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/efficient_keypoint_bbox_head.md)
 * [fomo_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/fomo_head.md)
 * [ghostfacenet_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/ghostfacenet_head.md)
 * [ocr_ctc_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/ocr_ctc_head.md)
 * [precision_bbox_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/precision_bbox_head.md)
 * [precision_seg_bbox_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/precision_seg_bbox_head.md)
 * [segmentation_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/segmentation_head.md)
 * [transformer_classification_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/transformer_classification_head.md)
 * [transformer_segmentation_head](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/heads/transformer_segmentation_head.md)
