Tags help you find and organize files, including images. With asset auto tagging enabled, you can also enable image auto tagging. Image auto tagging automatically tags images uploaded to the Documents and Media Library. This lets you use tags without requiring anyone to apply them manually.
Image auto tagging is disabled by default. To use it, you must do two things:
-
Ensure that asset auto tagging is enabled. Although it’s enabled by default, administrators can disable it.
-
Ensure that an image auto tagging provider is enabled. These providers are disabled by default. Here, you’ll learn how to enable/disable them.
Three such providers are available:
TensorFlow:
An open-source library that provides machine learning capabilities. TensorFlow
image auto-tagging in Liferay DXP is based on
TensorFlow’s LabelImage sample for Java,
and uses the Inception5h model. Use this with caution, since its accuracy is
limited.
Google Cloud Vision: Uses the Google Cloud Vision API to automatically tag images.
Microsoft Cognitive Services: Uses Microsoft Cognitive Services to automatically tag images.
Figure 1: The tags *freight car* and *electric locomotive* were automatically applied to this image.
Configuring TensorFlow Image Auto Tagging
Follow these steps to configure TensorFlow Image Auto Tagging:
-
Go to Control Panel → Configuration → System Settings → Assets.
-
Under VIRTUAL INSTANCE SCOPE, select TensorFlow Image Auto Tagging. The following settings are available:
Enable TensorFlow Image Auto Tagging: Check this box to select whether image auto-tagging is enabled by default in any instance that has asset auto tagging enabled. Note that you can override this value for specific instances, as the next section shows.
Confidence Threshold: TensorFlow assigns a confidence level between 0 and 1 for each tag, where 1 is the highest confidence and 0 is the lowest. This field sets the minimum confidence level that TensorFlow needs to apply a tag. Higher values yield fewer tags because TensorFlow needs more confidence before it applies a tag. Likewise, lower values yield more tags.
-
Click Save to save your changes.
You can override these settings for each instance from Control Panel → Configuration → Instance Settings → Assets.
Figure 2: Configure TensorFlow image auto-tagging for your portal instances.
To optimize performance, you can also control the process that runs TensorFlow image auto tagging:
-
Go to Control Panel → Configuration → System Settings → Assets.
-
Under SYSTEM SCOPE, select TensorFlow Image Auto Tagging Process. The following settings are available:
Maximum Number of Relaunches: The maximum number of times the process is allowed to crash before it is permanently disabled.
Maximum Number of Relaunches Time Interval: The time in seconds after which the counter is reset.
-
Click Save to save your changes.
Figure 3: You can fine tune the process that runs the TensorFlow image auto tagging in the portal.
Configuring Google Cloud Vision
Follow these steps to configure Google Cloud Vision image auto tagging:
-
Go to Control Panel → Configuration → System Settings → Assets.
-
Under VIRTUAL INSTANCE SCOPE, select Google Cloud Vision Image Auto Tagging. The following settings are available:
API Key: The API key to use for the Google Cloud Vision API. For more information, see Google’s documentation on API keys.
Enabled: Whether Google Cloud Vision image auto tagging is enabled.
-
Click Save to save your changes.
You can override these settings for each instance from Control Panel → Configuration → Instance Settings → Assets.
Figure 4: The Google Cloud Vision provider requires an API key.
Configuring Microsoft Cognitive Services
Follow these steps to configure Microsoft Cognitive Services image auto tagging:
-
Go to Control Panel → Configuration → System Settings → Assets.
-
Under VIRTUAL INSTANCE SCOPE, select Microsoft Cognitive Services Image Auto Tagging. The following settings are available:
API Key: Your API key for the Computer Vision API V2.
API Endpoint: The endpoint for the Computer Vision API V2 (e.g.,
https://westcentralus.api.cognitive.microsoft.com/vision/v2.0).Enabled: Whether Microsoft Cognitive Services image auto tagging is enabled.
For more information, see the Microsoft Cognitive Services documentation.
-
Click Save to save your changes.
You can override these settings for each instance from Control Panel → Configuration → Instance Settings → Assets.
Figure 5: The Microsoft Cognitive Services provider requires an API key and an endpoint.