# AI Capabilities

Luxonis devices run neural networks fully on-device for classification, detection, segmentation, and custom vision tasks—keeping
latency low and data private while freeing the host for control logic.

Across [RVC2](https://docs.luxonis.com/hardware/platform/rvc/rvc2.md) and
[RVC4](https://docs.luxonis.com/hardware/platform/rvc/rvc4.md), you can run networks with
[NeuralNetwork](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/neural_network.md) and
[DetectionNetwork](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/detection_network.md), deploy models in
[NNArchive](https://docs.luxonis.com/software-v3/ai-inference/nn-archive.md) or blob formats, and source them from [Model
Zoo](https://docs.luxonis.com/software-v3/ai-inference/model-source/zoo.md), [custom
training](https://docs.luxonis.com/software-v3/ai-inference/model-source/training.md), or integrations.

OAK cameras with [RVC2](https://docs.luxonis.com/hardware/platform/rvc/rvc2.md) deliver efficient on-device inference with
balanced performance for common detection, segmentation, and classification pipelines.

OAK4 devices with [RVC4](https://docs.luxonis.com/hardware/platform/rvc/rvc4.md) add more TOPS, faster preprocessing, and native
superblob/NNArchive support, enabling heavier backbones (e.g., YOLO-E, DepthAnythingV2, ...), multi-model pipelines on a single
device, and [NeuralDepth](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/neural_depth.md).

## AI building blocks

 * Run generic networks with
   [NeuralNetwork](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/neural_network.md) and
   hardware-accelerated
   [DetectionNetwork](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/detection_network.md)/[SpatialDetectionNetwork](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/spatial_detection_network.md).
 * Source models from [Model Zoo](https://docs.luxonis.com/software-v3/ai-inference/model-source/zoo.md), [custom
   training](https://docs.luxonis.com/software-v3/ai-inference/model-source/training.md), or
   [integrations](https://docs.luxonis.com/software-v3/ai-inference/integrations.md); convert with [Model
   Conversion](https://docs.luxonis.com/software-v3/ai-inference/conversion.md).
 * Package once with [NNArchive](https://docs.luxonis.com/software-v3/ai-inference/nn-archive.md); measure performance using
   [Benchmarking](https://docs.luxonis.com/software-v3/ai-inference/benchmarking.md).

## Typical use cases

 1. Object detection & tracking: run YOLO via
    [DetectionNetwork](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/detection_network.md) on RGB and feed
    IDs into [ObjectTracker](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/object_tracker.md) for
    persistent tracklets (optionally spatial with depth).
 2. Pose/landmarks & overlays: use [ParsingNeuralNetwork](https://docs.luxonis.com/software-v3/ai-inference/inference.md) which
    automatically parses the outputs in standard message types.
 3. Segmentation & fusion: run pixelwise models, then fuse with
    [depth](https://docs.luxonis.com/overview/toplevel-features/depth.md) or [spatial
    perception](https://docs.luxonis.com/overview/toplevel-features/spatial_perception.md) to measure volume, occupancy, or free
    space.

## Code-less quick start with App Store

### Data Collection App

On-device open-vocabulary detection (YOLOE via DepthAI) auto-collects snaps and lets the UI pick labels, set confidence, and
enable snap conditions.

[Open Data Collection App](https://hub.luxonis.com/control/store/a_KbitoLgrb93sybwMot3fgu?view=page)

## Guides and examples

### Generic NN Example

Run any neural model on OAK camera

[Generic NN Example](https://github.com/luxonis/oak-examples/tree/main/neural-networks/generic-example)

### Detection Network

Run Model Zoo YOLO on RGB

[Detection Network](https://docs.luxonis.com/software-v3/depthai/examples/detection_network/detection_network.md)

### Detection remap

Align detections between streams

[Detection remap](https://docs.luxonis.com/software-v3/depthai/examples/detection_network/detection_network_remap.md)

### Detection replay

Track detections on recorded video

[Detection replay](https://docs.luxonis.com/software-v3/depthai/examples/detection_network/detection_network_replay.md)

### Model Zoo example

Download and unpack HubAI models in-pipeline

[Model Zoo example](https://docs.luxonis.com/software-v3/depthai/examples/model_zoo/model_zoo.md)

### AI Inference Docs

Read more about inference in DepthAI

[AI Inference Docs](https://docs.luxonis.com/software-v3/ai-inference/inference.md)

### Need assistance?

Head over to [Discussion Forum](https://discuss.luxonis.com/) for technical support or any other questions you might have.