search

Jawabreh0 / HumanexAI-Masked-Face-Recognition

This repository contains the source code for a Face Recognition System using MTCNN for face detection, Facenet for feature extraction, and SVM for classification. The system is capable of recognizing the identity of a person even when they are wearing a mask.
MIT License
0 stars 0 forks source link

High unknown face rate #4

Open Jawabreh0 opened 1 year ago

Jawabreh0 commented 1 year ago

The algorithm updated to optimize the threshold, it's giving a high confidence rate but still giving a very high unknown face rate

Jawabreh0 commented 1 year ago

The high rate of "No Face Detected" messages produced by the algorithm is due to the face detection algorithm's ability to detect masked faces, and not a high rate of "unknown faces." "Unknown face" refers to the situation where the algorithm detects a face, extracts its features, but makes an incorrect prediction that it is an unknown face when it is actually known. On the other hand, a high rate of "No Face Detected" means that the MTCNN algorithm used for face detection is unable to detect the face at all. In the system pipeline, the first step is to detect the face, followed by feature extraction and face identification. Therefore, the failure is occurring at the first step of the pipeline, specifically the face detection step. The difference between "unknown face" and "No Face Detected" lies in the fact that the failure in the former case is related to the classification algorithm (SVM), whereas in the latter case, it is related to the face detection algorithm (MTCNN). Nevertheless, a new update is being developed to replace the MTCNN algorithm and resolve this issue.