Where: Channel | Metadata | General tab
Feature is only available in Color Factory Enterprise
Azure Computer Vision is a cloud service by Microsoft that uses machine learning (AI) to identify objects, faces and other artifacts in pictures.
Color Factory then stores this information in metadata so that users can search for the information and find relevant content.
What types of information can the Computer Vision service recognize?
- Tags and Keywords
- The accent and dominant color of images (Foreground/Background color, dowinant colors, whether a picture is B&W)
- Description - A description of an entire image in human-readable language, using complete sentences, e.g. "A grey dog running on a beach")
- Faces, gender and age - Gender, Age, whether an image is a portrait etc
- Celebrities - Celebrities' names are stored in the set metadata field
- Landmarks - Big Ben, the Eiffel Tower etc.
- Objects - Detects objects in images and stores their coordinates within the picture
- Brands - Detects known brands by their logo (currently only English language supported)
- Adult content - Detects whether images are likely to contain adult or racy content.
For each of these "vision types" there will be additional settings that let you control the precise information you want to record in the metadata.
Subscribing to Azure Computer Vision
Before you can use this feature, you need to have a valid subscription for the Azure Computer Vision service. If you need help navigating the Azure portal to locate it, see here.
How to configure automatic tagging
Click on the Azure Computer Vision button to open the configuration screen for this feature.
The Azure service that powers the image recognition requires a license. Start by filling in the API key based on the license information you have received.
Next, choose a region; this is the data center that powers the image recognition service. Different regions may offer slightly different image recognition services.
Adding types of information to recognize
Now, click on the Add button to add a "vision type" to recognize. The screenshot below shows the configuration of tags/keywords, where the service will try to extract keywords based on what it finds in the pictures processed in the channel.
The tags identified by the service are stored in the XMP field you set below.
The automatic confidence level works like this:
- If the image recognition returns three or less tags, all tags are added.
- If fewer than 10 tags are returned, tags with a confidence level of 40% or more are added to the picture.
- If more than 10 tags are returned, tags with a confidence level of 60% are added to the picture.
The confidence level can be set to automatic, or you can set a a fixed minimum confidence level manually by dragging the slider to set the level of confidence with which a keyword must be established for it to be added to the image. There is also the option to add the confidence level to the metadata.