Machine Learning and AI Functionalities in ZEN

You can train and use machine learning models as well as download and use dedicated AI models, like deep learning models for instance segmentation from arivis Cloud. With the necessary licenses, you can use these models in several modules, including the Image Analysis and Bio Applications.

Machine Learning Capabilities

You can import and export machine learning models for segmentation, denoising, and object classification via the respective tool on the Analysis tab. You can execute them with a dedicated function on the Processing tab (even without a license), or as part of an image analysis and/or Bio Application. To train such models directly in ZEN, you need the license for the AI Toolkit. Training and running machine learning models for segmentation and object classification requires an installation of the 3rd party Python Tools during the installation of ZEN. Training machine learning models for denoising requires an installation of Docker Desktop and a corresponding denoising container during the installation of ZEN. The machine learning functionality in ZEN includes:

Intellesis Segmentation to train and use models for segmenting images using pixel-classification, see Intellesis Segmentation.
Intellesis Denoising to train and use deep learning based models for the denoising of images with the Noise2Void (N2V) approach, see Intellesis Denoising.
Intellesis Object Classification to train and use models for classifying objects based on measured parameters of an analyzed image, see Intellesis Object Classification.

AI Models

Compatibility of Instance Segmentation Models

Instance segmentation models trained after Oct. 21st 2025 are only compatible with NVIDIA Graphics cards of newer generation with CUDA Compute Capability ≥ 7.0 (Volta or newer). If you want to use these newer models with older NVIDIA graphics cards, you can use the CPU fallback (under Tools > Options > arivis Cloud > Instance Segmentation deactivate the checkbox Enable NVIDIA GPU for Instance Segmentation).

Different functionalities in ZEN can also make use of trained AI models packaged as Docker containers. These functionalities therefore need the Docker Desktop software installed and running on your PC. For more information, refer to the installation guide and see Requirements for Docker Desktop. You can directly download AI models from arivis Cloud into ZEN, for which you need to establish a connection with an access token, see Creating and Entering an Access Token and Downloading AI Models. Additionally, certain modules use systemwide Docker containers that can be installed with the ZEISS Microscopy Installer during installation of ZEN.

Creating and Entering an Access Token

To be able to connect ZEN with arivis Cloud to download and use AI models, you need an access token. Take the following steps to create the token and enter it in ZEN.

Your PC has a connection to the internet and you are signed in on arivis Cloud.
You have started ZEN.

Click Tools > Options > arivis Cloud.
The options are displayed.
In your browser, open the access token page of arivis Cloud (https://www.arivis.cloud/app/user/client-keys).
The page for your personal access tokens is displayed.
Click New Access Token to create a new access token.
A new window opens.
In the New Access Token window, enter a name, an expiration date, and select the desired access scopes for the token. Then click Next.
Your personal access token is created and displayed.
Copy your Personal token. Optionally, you can also consider storing the key in a secure location in case it needs to be recovered.
Go back to ZEN to Tools > Options > arivis Cloud, paste the token into the input field, and click OK.

You have now created and entered an access token. You can now download and use AI models from arivis Cloud in ZEN.

Downloading AI Models

Powerful GPU

For AI models requiring a GPU for execution, the GPU needs to fulfil certain criteria, i.e. be powerful enough. For the use of AI models, the recommended hardware configuration is 8GB GPU. Only Nvidia GPUs and the CPU are supported. For further reference, also refer to the requirements of the Intellesis functionality, see Remarks and Additional Information.

In ZEN, you can download AI models from arivis Cloud, e.g. for instance segmentation. These models are packaged as Docker containers and therefore need the Docker Desktop software to work. Note that Docker containers from arivis Cloud are not shared between different Windows user accounts. Each user needs to re-download a model.

Docker Desktop is installed and running on your computer. For more information about Docker Desktop, see Requirements for Docker Desktop.
You have the model you want to download available on arivis Cloud and your PC is connected to the internet.
You have created an access token and entered it in ZEN, see Creating and Entering an Access Token.

Go to the Applications, open the AI Model Store tool and click Open AI Model Store.
The AI Model Store dialog opens to display all available AI models.
On the left side, click Download from arivis Cloud.
The model list is updated and displays all models on arivis Cloud that you have available.
Select the model you want to download in the list.
The Properties section on the right displays more detailed information about your selected model.
In the Properties section, select the model version you want to download and click Download Model.
The download of the model starts and the progress is displayed in the status bar. You have the possibility to cancel the download with the button next to the progress bar.

After a successful download, the model is displayed in the AI Model Store tool. It is now available in ZEN, even without an active internet connection.
You can now use the model, e.g. in the Automatic Segmentation step of the image analysis.

Importing AI Models