Edge AI Manager

This device is configuring…

The device is now updating its registration information, please hold on…

The device is waiting for response from the server, this will automatically timeout after 15 minutes.

If this takes more than a minute, you can restart the registration .

This device is not registered yet.

This device has not yet been registered. Please click the button below to start the registration process.

Choose or edit a device name, this will make it easier to identify the device later.
Debug status and settings

Check the status and the settings of the AI manager in the following tabs.



Model is running

Configuration details

Selected model

Please choose a model Please choose a model Multiple models selected

Number of cameras

No input configured

Output location

Output is disabled


Service Status
Browser Browser connection unknown.
Scailable API Online status unknown..
Module Online status unknown..
Output distributor Online status unknown..

Registration Details

not available

Model settings

First Model

Input settings

Only one input driver an be used at the same time.

Use the `settings.json` to change the socketname.

Camera 1

Refresh preview

Current input: Format:

Use v4l2-ctl --list-devices to see available cameras.

The v4l2 source looks like /dev/video1, check lsusb to see available cameras.

Use arv:// and an IP-address or device name.

The use the first autodetected aravis source do arv://auto.

Aravis defaults to RGB format if available and will fall back to grayscale otherwise.

To enforce BayerRGB, RGB or Grayscale, use respectively arv://bayer@ip arv://rgb@ip and arv://gray@ip.

For instance, for BayerRGB with autodetected aravis source: arv://bayer@auto.

Enter a URL with a static image. (JPG, PNG, TGA, BMP, GIF)
Input source settings
Optional camera authentication
Optional camera identification
Image resize and crop

Resize image to tensor: Image is fitted exactly to the tensor input dimensions without taking into account the original image's aspect ratio. As a result, the input tensor can be a "squashed" representation of the original input image.

Cut edges to fit: Image is fitted to the tensor input aspect ratio by cutting edges of the image either vertically or horizontally in case the image aspect ratio differs from the tensor aspect ratio… As a result, parts of the input image may not be part of the inference.

Pad edges to fit: Image is fitted to the tensor input aspect ratio by adding black borders either horizontally or vertically in case the image aspect ratio differs from the tensor aspect ratio.

Optional mask area
Optional line-crossing area

Model input routing

There are multiple active models and inputs. Please choose the models for each input.

Input: Camera 1

Output settings

Select the output protocol you would like to use.

The protocol used to transmit the inference results. Other options like modbus are also possible, contact us for more details.

REST output

The built-in output distributor sends results of the inference the Scailable dataviewer.
Data can be viewed on the device and in the Scailable. platform.

It is also possible to use a custom output endpoint.

For more information please check the documentation.

You can use a service like webhook.site or send data to other services with a REST interface like Elasticsearch or Node-RED.

The protocol used to transmit the inference results is REST by default. Other options like modbus are also possible, contact us for more details.

The formatting of the results is JSON by default.

Socket output

Socket output specifies the output socket to send the data to.

Global Options

Cloud connection mode
Switch between default online mode or always offline mode. Always offline mode is not recommended.
External trigger
When this is activated, inference will not be run until a signal is received externally. A signal can be sent by calling the /triggerInference endpoint.
Raw model output
The raw output for the model will be included or excluded based on this switch. Changing this might be needed when the visualisation does not show any inference results.
Output image storage

"Off" will not save the input image.
"Full size input" will save the input image.
"Model size input" will save the image as it is processed by the model.

When "Save all input images" is on, images will be saved with the following directory structure: [SaveImagesPath]/[%Y]/[%m]/[%d]

Model Options

Inference speed

Select a limit for the inferences in milliseconds, to save bandwidth and/or power. The model may not be fast enough to reach this maximum, but the module will not attempt faster inferences.

Trigger inference on change
To trigger a model inference only when there's a significant change between the previous and current frames.
Threshold of change in a pixel value to be considered a changed pixel.
Percentage of pixels in frame necessary to be changed above the pixel difference threshold to be considered a changed frame.
Model cut-off

The cut-off value (ranging between 0.0 and 1.0) determines whether an object is classified within a class.

Many models produce some kind of confidence score (between 0.0 and 1.0) which is cut-off to determine class membership.

Values close to 0.0 will produce more false positives, while values close to 1.0 will produce more false negatives.

Upload images with low certainty

If the threshold is 0 no images will be sent to the API.
If the threshold is higher than 0 the input image will be sent to the API when an inference result has a probability value below the threshold.
If there are more detected classes, any detection in the result, independent of the detected class that is below the threshold will trigger that the image is sent to the API.

Test configuration

This will start the Scailable module, take one image for each enabled camera, feed image(s) to your selected model, print the output as it will be sent to the selected output location, display the image(s) in the camera preview field(s), and stop the module again.

Run model


Online status unknown.. Online status unknown..

Password protection

The AI manager can be protected with HTTP Basic Authentication.
You should enable this if your device is publicly accessible, e.g. by the exposed port 8081.

Be careful with this option.
When you select the checkbox, you must log in with the chosen password for user "scailable" immediately.
If you forgot the password the only way to reset it is to delete and reinstall the module.

Check the status and the settings of the AI manager in the following tabs.