# backbones

Python API: `luxonis_train.nodes.backbones`

Backbones that turn an image into feature maps.

Most backbones return a list of feature maps. The stride increases from one map to the next. A neck or a head then reads the maps.
Some backbones differ:

 * [ContextSpatial](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/contextspatial.md)
   and
   [DDRNet](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/ddrnet/ddrnet.md)
   return features at 1/8 of the input size, for a segmentation head.
 * [DinoV3](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/dinov3/dinov3.md)
   returns a list with one CLS embedding instead of feature maps when `return_sequence` is `True`.
 * [GhostFaceNet](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/ghostfacenet/ghostfacenet.md)
   targets face embeddings.
 * [RecSubNet](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/recsubnet/recsubnet.md)
   returns a packet with an image reconstruction and the original input, for anomaly detection.

A backbone with variants takes `variant` to pick a size. Each backbone docstring lists its variants and the parameters that each
variant sets.

## Child Pages

 * [ddrnet](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/ddrnet.md)
 * [dinov3](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/dinov3.md)
 * [efficientrep](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/efficientrep.md)
 * [efficientvit](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/efficientvit.md)
 * [ghostfacenet](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/ghostfacenet.md)
 * [micronet](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/micronet.md)
 * [mobileone](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/mobileone.md)
 * [pplcnet_v3](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/pplcnet_v3.md)
 * [recsubnet](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/recsubnet.md)
 * [repvgg](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/repvgg.md)
 * [contextspatial](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/contextspatial.md)
 * [efficientnet](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/efficientnet.md)
 * [mobilenetv2](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/mobilenetv2.md)
 * [resnet](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/resnet.md)
 * [rexnetv1](https://docs.luxonis.com/software-v3/ai-inference/model-source/training/luxonis-train/luxonis-train-api-reference/nodes/backbones/rexnetv1.md)