# losses

Python API: `luxonis_train.attached_modules.losses`

损失函数，按其监督的任务分组。

 * 分类和分割：[CrossEntropyLoss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/cross_entropy.md)、[BCEWithLogitsLoss](https://docs.luxonis.com/zh/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/bce_with_logits.md)、[SmoothBCEWithLogitsLoss](https://docs.luxonis.com/zh/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/smooth_bce_with_logits.md)、[SigmoidFocalLoss](https://docs.luxonis.com/zh/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/sigmoid_focal_loss.md)、[SoftmaxFocalLoss](https://docs.luxonis.com/zh/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/softmax_focal_loss.md)
   和
   [OHEMLoss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/ohem_loss.md)
 * 边界框：[AdaptiveDetectionLoss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/adaptive_detection_loss.md)
   和
   [PrecisionDFLDetectionLoss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/precision_dfl_detection_loss.md)
 * 实例关键点：[EfficientKeypointBBoxLoss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/efficient_keypoint_bbox_loss.md)
 * FOMO
   对象中心：[FOMOLocalizationLoss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/fomo_localization_loss.md)
 * 实例分割：[PrecisionDFLSegmentationLoss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/precision_dfl_segmentation_loss.md)
 * 异常检测：[ReconstructionSegmentationLoss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/reconstruction_segmentation_loss.md)
 * OCR：[CTCLoss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/ctc_loss.md)
 * 嵌入：[pytorch-metric-learning](https://kevinmusgrave.github.io/pytorch-metric-learning/losses/) 损失，由
   [EmbeddingLossWrapper](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/embedding_losses.md)
   封装

配置将损失附加到节点的 `losses`
列表中的节点。每个损失的“兼容”部分列出了它接受的节点。[BaseLoss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/base_loss.md)
描述了如何编写新的损失。当没有节点有损失时，训练步骤会引发 `ValueError`。

## 子页面

 * [adaptive_detection_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/adaptive_detection_loss.md)
 * [base_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/base_loss.md)
 * [bce_with_logits](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/bce_with_logits.md)
 * [cross_entropy](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/cross_entropy.md)
 * [ctc_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/ctc_loss.md)
 * [efficient_keypoint_bbox_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/efficient_keypoint_bbox_loss.md)
 * [embedding_losses](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/embedding_losses.md)
 * [fomo_localization_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/fomo_localization_loss.md)
 * [ohem_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/ohem_loss.md)
 * [precision_dfl_detection_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/precision_dfl_detection_loss.md)
 * [precision_dfl_segmentation_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/precision_dfl_segmentation_loss.md)
 * [reconstruction_segmentation_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/reconstruction_segmentation_loss.md)
 * [sigmoid_focal_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/sigmoid_focal_loss.md)
 * [smooth_bce_with_logits](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/smooth_bce_with_logits.md)
 * [softmax_focal_loss](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/attached_modules/losses/softmax_focal_loss.md)
