Amazon Rekognition Overview
What is Amazon Rekognition?
Rekognition is a service used to find objects, people, texts, scenes in images and videos using machine learning.
What can Amazon Rekognition do?
So it can do facial analysis and facial search to do user verification and count how many people there are in an image, for example, or you can create your own database of familiar faces or you can compare against a database of celebrities to find people in your images as well.
Use Cases for Rekognition
So the use cases for Rekognition is to take your image and do labeling images and video, content moderation, text detection, face detection and analysis, for example, find the gender, the age range and the emotions that are associated with these faces, do a face search and verification, do celebrity recognition, as well as pathing.
So for example, if you’re doing a sports game analysis of a video, so that’s a high level.
Remember Rekognition is about images and videos.
How does that work?
Well, the image will be analyzed by Amazon Rekognition and then you set a Minimum Confidence Threshold for items that will be flagged.
So you set it to whatever percentage you want, and obviously the lower you set this percentage the more matches you’re going to get.
This confidence percentage represents how confident Amazon Rekognition is that this flagged image represents indeed some inappropriate or offensive character, and then once you have done thisand you flagged some images, you may want to do a human manual review, and so to do so, you can use something called Amazon Augmented AI (A2I) and you do optional manual review directly in Amazon A2I.
All of this process allows you to automatically flag images that can be sensitive and then use a final manual review to know whether or not you want to keep them or delete them.
This can help you with comply with regulations in case you must detect these kind of content before they’re posted to your applications.
