Using all the building blocks discussed until now, we will build an image recognition-based building access management system. This system has three main components.

The first is creating the collection, which will store the feature information of all the images.

The second is where the image of a new person is uploaded for the first time in the face collection. This task is executed automatically when the person stands in front of the IR sensor (interfaced with Raspberry Pi). The sensor detects their presence and triggers the camera to capture the image of their face. Then, the image is uploaded in S3 and the IndexFaces operation is performed, which extracts the feature information from the face and stores it in a collection.

In the third component, when a person stands near the entrance, an IR sensor detects their presence and triggers the camera to capture the image of the standing person. Then, the captured picture is compared with the reference images already stored in the system, using the SearchFacesByImage method. It compares the face by extracting the features of the face in the image and if the two faces match, it allows the person to enter the building (glow LED), and if not, it raises an alert.

Figure 7.21 shows the architecture of our implementation:

Let's begin with the implementation step by step.