# metrics

Python API: `luxonis_train.attached_modules.metrics`

比较节点预测与标签的指标。

指标通过配置中节点的 `metrics` 列表附加到节点。训练器在每个验证和测试批次上更新每个指标。在 epoch 结束时，它计算并重置每个指标并记录结果。

在配置中用 `is_main_metric` 标记一个指标。如果没有指标设置它，配置会标记第一个指标。训练器将主指标最高值的检查点保存在 `best_val_metric` 目录中。

[MeanAveragePrecision](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/mean_average_precision/mean_average_precision.md)
和
[DetectionConfusionMatrix](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/confusion_matrix/detection_confusion_matrix.md)
在头部非极大值抑制后使用边界框，因此其 `conf_thres` 和 `iou_thres`
会影响其结果。[PrecisionRecallCurve](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/precision_recall_curve.md)
使用原始边界框并执行自己的抑制。

要编写新指标，请继承
[BaseMetric](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/base_metric.md)。[MetricState](https://docs.luxonis.com/zh/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/base_metric.md)
的示例展示了一个完整的子类。

## 子页面

 * [confusion_matrix](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/confusion_matrix.md)
 * [mean_average_precision](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/mean_average_precision.md)
 * [base_metric](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/base_metric.md)
 * [dice_coefficient](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/dice_coefficient.md)
 * [embedding_metrics](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/embedding_metrics.md)
 * [mean_iou](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/mean_iou.md)
 * [object_keypoint_similarity](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/object_keypoint_similarity.md)
 * [ocr_accuracy](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/ocr_accuracy.md)
 * [precision_recall_curve](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/precision_recall_curve.md)
 * [torchmetrics](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/torchmetrics.md)
 * [utils](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/metrics/utils.md)
