Launched in 2016, AWS Rekognition is a cloud based SaaS (Software as a Service) facial recognition platform and developer tool provided by Amazon for developers working with Amazon Web Service, allowing them to conveniently incorporate identification of various objects, including recognizing people, objects, scenes and activities in images and videos to their applications with fast-paced, scalable technology. This service also provides the convenience of recognizing any inappropriate content.
AWS Rekognition is a product of the artificial intelligence and machine learning category and incorporates deep learning in order to detect objects in images and videos. It does not require developers to have knowledge of machine or deep learning, and is extremely useful in various aspects including unlocking smartphones, identifying profiles on social media, diagnosing illnesses, and finding missing persons.
AWS Rekognition provides precise facial recognition on both images and videos. It can enable users to conduct background checks, count people and carry out other various functions according to the requirements of their business. Through Rekognition Video, developers can also track an object’s path in a video, through a process known as pathing.
Privacy and loyalty of customers is the highest priority of Amazon Web Services; they do not use any personal information for marketing purposes. The security department at Amazon ensures that proper controls are in place in order to prevent any unauthorized and unauthenticated access. Furthermore, no faces are saved in Amazon’s database; the only storage that takes place is of the vectors of the face in order to compare it with other faces.
Amazon utilizes deep neural networks to identify and label millions of objects and people in images and videos and is continuously working on the creation and addition of new labels and features to make Rekognition even more essential for its users.
Currently, Amazon supports JPEG and PNG formats for images and MPEG-4 and MOV for videos. For videos, Amazon further puts a condition for them to be encoded with the H.264 codec, which is the most popular choice for recording and compressing videos. As far as sizes are concerned, Rekognition allows for an image size of up to 15 MB when submitted as an S3 object and caps the limit at 5 MB when it is submitted as a byte array. For videos, the limit is capped at 10 GB, allowing for approximately 6 hour videos.
Features of AWS
There are several features of AWS that allow it to identify and recognize different aspects of images and videos:
Through AWS, developers can identify various objects, scenes and activities in a photo or video such as a car, an ocean, a landscape or people playing soccer. Bounding boxes can be used to find the exact location of objects or people in a picture as well as enabling developers to get exact measurements of the object if required.
Furthermore, AWS Rekognition also provides a percentage of the accuracy of the detected label by using the SearchFaces or DetectLabels function. This percentage is known as the confidence threshold, which is set at a default score of 50.
Amazon Rekognition Custom Labels, a more specific version of Labels, can identify images and texts that are specifically required for your business. For example, detecting your products on various websites or identifying your logo on social media sites. Developers will add a specific set of images as training images which will then be used to compare and produce a custom image analysis for the client.
The “pathing” function of AWS Rekognition allows for the tracking of an object or a person in a video shot. While Amazon has clearly mentioned that pathing cannot be tracked across multiple cameras and is unable to combine with facial recognition, it is however possible on single camera-shots.
Using AWS Image and Video Moderation API, it is easy to detect, control and hide inappropriate or unsafe content on both images and video based on the requirements of your business. This includes graphic content, partial and full nudity and more.
Using the text detection feature in AWS Rekognition enables you to detect text in both images and videos which is then converted into machine readable text which can then be used for both .jpeg and .png images. DetectText and GetTextDetection can identify upto 50 words per frame in image and video respectively and can also identify numbers and symbols.
Through the AWS Face Recognition feature, users can identify faces in images and videos, with information including their face dimensions as well as the emotions and sentiments that are projected by the face. The facial and sentiment analysis can prove to be useful for corporate businesses and industries. Moreover, Rekognition allows facial comparison in different images and videos. However, this feature is only applicable for human faces and does not include cartoons, animated characters or non-human creatures.
AWS collects and stores metadata regarding face detection in containers called collections. Every face collection has their own specific ARM (Amazon Resource Name) with the most recent face identification on priority. This stored information can assist users in searching and comparing for faces in images and videos, using the SearchFaces and CompareFaces API functionality.
AWS Rekognition has a celebrity recognition API which enables facial recognition of thousands of celebrities in different industries. It identifies celebrities in different environments, with and without makeup, in different wardrobes and on different sets.
Other features of AWS Rekognition include:
Searchable image library
Face based user verification
How Does AWS Rekognition Work?
AWS Rekognition has 2 sets of API, one for images and the other for videos. Both APIs are used for recognition analysis to add observation and insight into the user’s application. It compares the photo in an image in the database with the image to be identified and detects a match with the use of the primary functions of indexing and searching faces.
Rekognition works by assigning face IDs to every image. It captures a live image, transforms it into face ID and carries out the search in the database for similar faces and returns a match, adding a percentage of confidence based on the feature matching.
Benefits of AWS Rekognition
AWS allows for easy and quick integration of image detection and recognition features into the user’s application, making it a time efficient choice for facial recognition functionality.
Scalable Image Recognition
AWS Rekognition can analyze millions of photos within a short span of time and can provide the user with results regardless of the amount of requests that are provided to it. Basically, this means that it doesn’t matter whether you send in one request or thousands of them; the response time will remain the same.
Integration with AWS Components
Rekognition is created to easily integrate with other components of AWS including Amazon S3, AWS Lambda and other such services.
There are no initial payments or minimum payment required by AWS Rekognition services. The only cost that developers will need to endure is for the quality of images that they want to analyze and for the metadata for faces that they will store.
Why Choose Winterwind For AWS Rekognition Integration?
Many corporations and industries find the AWS Rekognition tool useful; social media organizations use it to eliminate fake profiles, other corporate organizations have used it to label individuals in photos and videos. Financial companies are choosing it to authenticate and validate transactions. Globally, people want to develop applications which are able to detect and identify objects, faces and landscapes.
We can provide the following services for AWS Rekognition:
- Integration with app for mobile and other devices
- Improved facial and sentiment analysis
- Extensiveness of the tool to incorporate analysis of millions of images in a short span of time
- Integration and support with other AWS components
Doppelgang is an application that searches for celebrity look-alike faces for the user. Users can log in to the app, take a photo of themselves using the camera or use a picture from the camera roll, and the app searches in the database for the celebrity that most resembles the person in the photo.
In order for us to build this app, we first created a collection to organize our list of indexed faces of the celebrities which enabled the app to determine which collection to look at when searching for the user’s look-alike. In just a matter of seconds, the app can recognize the face or faces in the image using index faces.
When a user selects a photo using Doppelgang, the app will first detect the faces in the image. Amazon Rekognition will be able to detect the face. The app will then use the search faces function, searching faces in the collection of indexed faces to compare if the faces resemble any of the faces of the celebrity. The user will be able to see the result of which celebrity is their look-alike with a confidence percentage along with it.
Rekognition played a major role in enabling us to seamlessly build the Doppelgang application. We did not need to worry about the process of how the app will detect, search and compare for faces, or learn about machine learning. We were able to build the app quickly, which allowed for better customer satisfaction as well.
Ready for your next project using Amazon Rekognition? Talk to our experts today or send us an email! We would love to have a discussion with you!