I an working on a project of face detection and recognition, our goal is to detect the face from the image(frame) which is the output of the front-end socket and stored in a folder, my part for this project is to get that image and detect the face first and then try to see whether that face has came for the first time or it is a known face. The comparison algorithm is compare_faces inside face_recognition library which is available in the python. I am using MTCNN for detecting faces, for the landmarks I am using the same MTCNN which is pretty accurate but the problem comes in the blur faces. Even though the faces are blur my program passes it from all the constraint and let that face be a unique face even though the same face with the clear image is already stored in the known_faces variable. The difference between image with same face but different clarity is visible with naked eye but i just want to skip the blur face image so that my program works properly. How can I skip the face image which are blur
My main problems are blur face images, I want to skip them as much as possible. I tried Laplacian variable, FFT and many things but they are not accurate because the project is not just a college project, it is a production level project so we can not allow this errors.