necessary preprocessing steps to improve face verification performance

serengil / deepface

A Lightweight Face Recognition and Facial Attribute Analysis (Age, Gender, Emotion and Race) Library for Python

https://www.youtube.com/watch?v=WnUVYQP4h44&list=PLsS_1RYmYQQFdWqxQggXHynP1rqaYXv_E&index=1

MIT License

14.34k stars 2.2k forks source link

necessary preprocessing steps to improve face verification performance #407

Closed sri9s closed 2 years ago

sri9s commented 2 years ago

As I use face verification to test on a new dataset I finetuned the threshold and got the best model and distance metric.

Can you also guide in any preprocessing steps that is to be taken to further improve the model performance?

serengil commented 2 years ago

detection and alignment are pre-processing stages. you can use many different face detectors within deepface: opencv, ssd, mtcnn, dlib, retinaface, mediapipe.

find the best detectors for your case. I mean that retinaface might be the most accurate but it is very slow. On the other hand, opencv is the fastest but its alignment score is very low.