Model Zoo

DepthAI Model Zoo is a collection of neural network models deployable to OAK and RAE. You can access the full list of models directly in our GitHub repository. For licensing, please refer to the information in model.yml in the model's respective directory.

How to use

There are different ways to access the models from the DepthAI model zoo.

Note that the models in the DepthAI Model Zoo are exported such that they expect images with values in the [0-255] range, BGR (Blue Green Red) color order, and CHW (Channel Height Width) channel layout.

Python (DepthAI)

To use seamlessly with DepthAI, you can use our BlobConverter library:

Install BlobConverter using pip:

Command Line

1pip install blobconverter

Call from_zoo method with zoo_type="depthai", the name of the model that you want, and a number of shaves to use:

Python

1import blobconverter
2model_path = blobconverter.from_zoo(
3        name="yolop_320x320",
4        zoo_type="depthai",
5        shaves=6
6    )

It is also possible to download models from Open Model Zoo by setting zoo_type="intel". And by calling blobconverter.zoo_list(zoo_type=...) you can get the list of all models available in the chosen zoo.

Pass the model_path to the neural network node:

Python

1neural_network.setBlobPath(model_path)

The default path for the downloaded blob is the ~/.cache/blobconverter directory.

It is arbitrary and can be specified by setting the output_dir parameter.

BlobConverter interface

You have the option to download each exported model either through our BlobConverter web interface or by utilizing the BlobConverter API. Please note that downloading from the interface is only available from version 2022.1. You can find additional information about the conversion process here.

ON THIS PAGE

Model Zoo

How to use

Python (DepthAI)

BlobConverter interface