using aws rekognition for face detection and comparision

Question

I am trying to implement a face rekognition service that will support grouping images based on faces.
The problem is that the faces found by indexFaces are not high-quality enough for comparison ( serachFaces , searchFacesByImage) even when I use the HIGH quality filter in indexFaces.
Is there a way to index the faces in a dataset in a way that will ensure that each face found would be matched with a high certainty to another face if they are of the same person? 
The problem I encounter is:
I run indexFaces which detects and indexes 1000 faces.
Then I iterate face by face and use SearchFaces to find all the other photos in which this specific face is in
Some portion of the faces ~ 15% have no match but they are clearly present in other images just not comparable by searchImages

Pedro Correia · Answer

Hi there,

To ensure higher certainty in matching faces of the same person, you can try the following approaches:

1. Improve image quality: Ensure that the images you provide for face detection and comparison are of high quality, well-lit, and have clear visibility of the faces.

2. Adjust the similarity threshold: When using the SearchFaces or SearchFacesByImage API, you can adjust the similarity threshold parameter to control the sensitivity of the face comparison. Setting a higher threshold may result in more accurate matches but could also increase the chance of false negatives.

3. Consider using facial landmarks: AWS Rekognition provides APIs to detect facial landmarks such as eyes, nose, and mouth. Utilizing this information along with face detection can help improve the accuracy of face comparison.

4. Use additional face recognition algorithms: You can explore other face recognition algorithms or libraries that may provide better results for your specific use case. OpenCV and dlib are popular alternatives that you can consider.

If you need further assistance or have specific questions regarding the AWS Rekognition service, I recommend checking the help section located in the persistent menu or contacting AWS support for more detailed guidance tailored to your specific scenario.

Best regards

using aws rekognition for face detection and comparision

Relevant content