Amazon Rekognition

With the Amazon Rekognition plugin, you can quickly associate search terms with images and videos. Editors can then search for and use your media in their own content.

Configuring Amazon Rekognition

Administrators perform this task.

To configure Amazon Rekognition:

  1. From the Navigation menu, select Admin > Sites & Settings.

  2. In the Sites widget, click Global. The Edit Global widget appears.

  3. (Optional) Under Main, expand AWS Rekognition. In the Region field, enter the region to use when making requests to AWS Rekognition (or leave blank to use the region in your Beam configuration).

    ../../../_images/aws-region.png
  4. Under Main, expand AWS Rekognition (Image), and configure the confidence values for image recognition. Generally, higher confidence levels provide more accurate results (fewer false positives) but may miss some matches (more false negatives).

    ../../../_images/rekognition_config_image.png
  5. Expand AWS Rekognition (Video), and make the following settings:

    • In the Time Frame Interval (Seconds) field, enter the number of seconds by which Brightspot groups Rekognition’s results (see the illustration Video recognition results grouped by time frame for an example). If you leave this field blank, Brightspot groups the results in 60-second intervals.
    • Configure the confidence values for video recognition. Generally, higher confidence levels provide more accurate results (fewer false positives) but may miss some matches (more false negatives).
    • Configure the other fields corresponding to your account settings with AWS.
    ../../../_images/rekognition_config_video.png

    Configuring AWS Rekognition for video

  6. Click Save.

Rekognition is configured, and editors can view the results of a Rekognition analysis in the content edit form.

Submitting an image or video to Amazon Rekognition

You can use Amazon’s Rekognition service for associating relevant search terms to images or videos you upload to Brightspot.

To submit an image or video to Amazon Rekognition:

  1. Create a new image or video.
  2. Under the Metatdata tab, review the analysis returned from Rekognition.
../../../_images/rekognition-input.jpg

Image submitted to Amazon Rekognition

../../../_images/rekognition-result.png

Result from Amazon Rekognition

When editors search for keywords in the Search Panel related to the image, Brightspot includes the submitted image or video in the search results. Referring to the previous example, if an editor searches for Cat Manx in the Search Panel, Brightspot includes the above image in the search results.

If your connection to Amazon Video Rekognition is configured to group results by time frame (see illustration Configuring AWS Rekognition for video), Brightspot groups the video recognition results accordingly.

../../../_images/aws-rekognition-by-time-frame.png

Video recognition results grouped by time frame

Referring to the previous illustration, within the interval 45 seconds to 1 minute, Rekognition detected a ball, water, a heart, and a play area in the video.