# predefined_models

Python API: `luxonis_train.config.predefined_models`

根据名称和少量参数构建的模型图。

在配置的 `model.predefined_model` 部分中命名模型。然后，配置将模型的节点附加到
`model.nodes`，位于配置列出的节点之后。这些节点承载模型的损失、指标和可视化工具。`include_losses`、`include_metrics` 和 `include_visualizers` 控制节点保留哪些。

大多数模型提供 `light` 和 `heavy` 变体，有些模型在它们之间提供 `medium` 变体。`light` 最快，`heavy` 最准确。每个模型文档字符串都列出了其组件和变体，大多数还列出了测量的吞吐量。

预定义模型是版本化的。`version` 选择版本，`"latest"` 是默认值。更改现有模型图的更改将进入新的版本目录。因此，固定 `version` 的配置仍会构建相同的图。

## 子页面

 * [anomaly_detection](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/config/predefined_models/anomaly_detection.md)
 * [classification](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/config/predefined_models/classification.md)
 * [detection](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/config/predefined_models/detection.md)
 * [embeddings](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/config/predefined_models/embeddings.md)
 * [fomo](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/config/predefined_models/fomo.md)
 * [instance_segmentation](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/config/predefined_models/instance_segmentation.md)
 * [keypoint_detection](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/config/predefined_models/keypoint_detection.md)
 * [ocr_recognition](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/config/predefined_models/ocr_recognition.md)
 * [segmentation](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/config/predefined_models/segmentation.md)
 * [base_predefined_model](/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/config/predefined_models/base_predefined_model.md)
