Does Amazon Rekognition just need one image to recognise faces?

Question

I am creating a collection using the amazon rekognition create collection api call . Does each person need only one image for him to be classified well? Or do we need to give multiple images per class(person) as done in facenet or other deep learning implementations to extract features ?

I have already added all the images(multiple images per person) and it shows me it has detected someone well enough. But can the collection cluster similar featured images to form one person ?

Shailendra Shailendra · Accepted Answer · 2019-05-01T15:28:05

Adding faces to collection is done by IndexFaces operation which actually detects faces and add them to collection. For each face - it will return you faceid and other face details.

{
    "FaceModelVersion": "3.0",
    "FaceRecords": [
        {
            "Face": {
                "BoundingBox": {
                    "Height": 0.3247932195663452,
                    "Left": 0.5055555701255798,
                    "Top": 0.2743072211742401,
                    "Width": 0.21444444358348846
                },
                "Confidence": 99.99998474121094,
                "ExternalImageId": "input.jpg",
                "FaceId": "b86e2392-9da1-459b-af68-49118dc16f87",
                "ImageId": "09f43d92-02b6-5cea-8fbd-9f187db2050d"
            },
            "FaceDetail": {
                "BoundingBox": {
                    "Height": 0.3247932195663452,
                    "Left": 0.5055555701255798,
                    "Top": 0.2743072211742401,
                    "Width": 0.21444444358348846
                },
                "Confidence": 99.99998474121094,
........................
.........................
........................

For each face detected, Amazon Rekognition extracts facial features and stores the feature information in a database. In addition, the command stores metadata for each face that's detected in the specified face collection. Amazon Rekognition doesn't store the actual image bytes.

This faceid is sufficient to search on any collections containing faces using SearchFaces operation. The operation compares the features of the input face with faces in the specified collection. Of course there is a degree of match or similarity which is happening to arrive at the match.

The operation response returns an array of faces that match, ordered by similarity score with the highest similarity first. More specifically, it is an array of metadata for each face match that is found. Along with the metadata, the response also includes a confidence value for each face match, indicating the confidence that the specific face matches the input face.

Does Amazon Rekognition just need one image to recognise faces?

1 Answers