Visual Recognition applies image recognition algorithms to your content to detect relevant features and automatically apply "Concepts".

Concepts are similar to Tags however they are simpler, text only labels and can be controlled and managed separately from Tags so as not to complicate your Tagging strategy.

How to configure Visual Recognition

Navigate to Plugins > Visual Recognition > Configure.

There is a single configuration screen for Visual Recognition, all your available options are found here.

visual-recognition-configure.jpg

Automatically apply to selected terms

From this dropdown you can select which one or more of your Terms you would like to apply Visual Recognition to.

NB: You can also choose to apply Visual Recognition to a Term when creating or editing a Term, you'll find the option on the Advanced tab.

Only apply concepts with ≥ X% confidence

As the description states, each Concept that Visual Recognition identifies comes with an estimate of how accurate it is - this is called "confidence".

Choosing a higher confidence level eg. 95% will ensure the Concepts applied to your content are more accurate but typically you will receive fewer. Lowering the confidence will result in more Concepts being applied however the accuracy of some may be lower.

You can test different confidence levels by selecting from the dropdown and then observing the results in the Concepts Preview window below.

Rate Limit

The information found here shows you how many images you are able to apply Visual Recognition to each month, how many you have used and how many days are left until your monthly quota is refreshed.

Advanced Visual Recognition Additional Options

If you have access to advanced Visual Recognition you will see additional options.

visual-recognition-configure-advanced.jpg

Video Processing

This option applies Visual Recognition to Instagram, Twitter and GoConnect uploaded videos.

Advanced Visual Recognition Models

Advanced Visual Recognition provides you with access to a number of speciality models that have been trained for specific types of content.

You can read more about these models here.

Did this answer your question?