seasonSH / DocFace

Face recognition system for ID photos
MIT License
371 stars 124 forks source link

How to perform the final verification/classification? #1

Closed firmamentqj closed 6 years ago

firmamentqj commented 6 years ago

Hi,

The guideline has provided the way to extract features for test images. But how to verify if two images belong to the same person or not? To calculate the distance between the extracted features? Can the pre-trained/fine-tuned network be used to output the probability of that?

Many thanks!

seasonSH commented 6 years ago

You can just use the Euclidean distance for comparing the extracted features.

firmamentqj commented 6 years ago

Thanks for your reply!

But how can I decide whether two images belong to the same person or not based on their Euclidean distance? In other words, how to determine the distance/similarity threshold?

seasonSH commented 6 years ago

You need to decide the threshold using a validation dataset. For example, you can choose a threshold where the False Accept Rate equals 0.1% for pairwise verification.