At the AWS re:Invent conference, Amazon launched Rekognition, a managed service for Image Recognition and Analysis, powered by Deep Learning. The capabilities that Rekognition provides include Object and Scene detection, Facial Analysis, Face Comparison and Facial Recognition.
Image Source: (screenshot) https://www.youtube.com/watch?v=fk-TxySUAzw
Image growth has been on the rise over the past five years, with 1.2 Trillion photos expected to be taken in 2017, as opposed to the 80 billion photos estimated to have been taken in 2000. The 2016 Mary Meeker Internet trends report estimates that 1 billion photos are taken every day on Snapchat alone.
Image Source: (screenshot) https://www.youtube.com/watch?v=fk-TxySUAzw
As more and more content moves to digital images, how can organizations index and interpret this data? Ranju Das, general manager for Amazon Rekognition, positions the service as a way "to find meaning in your data." While there are many services available that can provide insight into your data, with Reckognition, "Amazon has tried to solve that problem for visual content."
While Amazon Rekognition is a new public service, it has a proven track record. Jeff Barr, chief evangelist at AWS, explains:
Powered by deep learning and built by our Computer Vision team over the course of many years, this fully-managed service already analyzes billions of images daily. It has been trained on thousands of objects and scenes. Rekognition was designed from the get-go to run at scale.
For organizations looking to explore the service, they can try the console for quick demo scenarios before writing code and consuming the Rekognition APIs.
Object and Scene Detection
Object and Scene Detection can be used to identify objects within an image. For example, take a Living Room that has many objects in it. Using Object and Scene Detection, Amazon Rekognition will be able to identify common objects like Flowers, Coffee Tables, and Chairs.
Image Source: (screenshot) https://www.youtube.com/watch?v=fk-TxySUAzw
Amazon has identified the following use cases for Object and Scene Detection:
- Photo-sharing apps can power smart searches and quickly find cherished memories, such as weddings, hiking or sunsets.
- Vacation rental markets can automatically label host-uploaded images with tags such as fireplace, kitchen or swimming pool.
- Travel sites and forums can classify user generated images with labels such as beach, camping or mountains.
Amazon has been using Rekognition within its Amazon Prime Photo service, which provides free photo storage. Amazon has indicated that “Billions of photos” have been uploaded to this service. As a result, Amazon can provide metadata about the content of the photos that have been uploaded.
Facial Analysis
Using Facial Analysis, developers have access to characteristics of a photo including:
- Demographic Data
- Sentiment Expressed
- Facial Landmarks
- Image Quality
- General Attributes i.e. whether someone has their mouth open
Image Source: (screenshot) https://www.youtube.com/watch?v=fk-TxySUAzw
Some of the use cases that Amazon has identified for Facial Analysis include:
- Photo printing service can recommend the best photos to users.
- Online dating applications can improve their match recommendations using facial attributes.
- Ad-tech services can display dynamic and personalized content to customers.
Another use case that Amazon has identified is in the area of Retail customer service. Consider a scenario where you may have shoppers in a store looking for goods. Using Facial Analysis, a store could determine a customer’s sentiment and pro-actively initiate contact before a customer decides to leave as a result of frustration.
Image Source: (screenshot) https://www.youtube.com/watch?v=fk-TxySUAzw
Face Comparison
Amazon describes the Face Comparison API as a way to:
Measure the likelihood that faces in two images are of the same person. With Rekognition, you can use the similarity score to verify a user against a reference photo in near real time.
Image Source: https://aws.amazon.com/rekognition/
The Face Comparison API depends upon a Similarity attribute that can be used to drive different behaviors within an application that is using it. For example, consider an application that is using Face Comparison for Security purposes: to enter a building. You would want to have this Similarity threshold to be much higher than an application that will change the brightness of lights based upon someone entering a room.
Other use cases for Face Comparison may include:
- Surveillance scenarios where you may be looking for people of interest or concern.
- IoT and device manufacturers can build face-based verification directly into their applications.
- Hospitality businesses may provide additional customer service for VIPs once they have been identified.
Face Recognition
Allows you to build an index of images that then can be used for comparison purposes. When this occurs, Amazon is using the metadata representation of the image in the search.
Image Source: https://aws.amazon.com/rekognition/
Amazon Rekognition is available in US East (Northern Virginia), US West (Oregon), and EU (Ireland) Regions. There is a Free Tier that allows developers to analyze up to 5000 images per month and save 1000 face vectors each month for an entire year. There is consumption-based billing for additional usage.
