You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, David.
Can you tell me why you set the threshold as 0.6 and the percentage threshold as 70?
I knew the "threshold" in FaceNet which is used to distinguish faces are from the same person or not, but it seems that the "threshold" in your program is different.
Can you tell me why you set two different thresholds?
And one more question.
I saw the equation in your source code main.py :
percentage = min(100, 100 * thres / smallest)
I downloaded your source code and it worked well for me.
I wonder why the real-time face recognition can work well by using this equation.
Is there something like principle to explain it?
Looking forward to your reply. :)
The text was updated successfully, but these errors were encountered:
I think the real equation for checking the similiarity of face is the ecluidean distance, that check similiarity vector from realtime face recognition and with the vector data of our trained face before.
That equation "percentage = min(100, 100 * thres / smallest)" is for showing percentage of similiarity that system recognize the face based on our trained data.
Hi, David.
Can you tell me why you set the threshold as 0.6 and the percentage threshold as 70?
I knew the "threshold" in FaceNet which is used to distinguish faces are from the same person or not, but it seems that the "threshold" in your program is different.
Can you tell me why you set two different thresholds?
And one more question.
I saw the equation in your source code main.py :
percentage = min(100, 100 * thres / smallest)
I downloaded your source code and it worked well for me.
I wonder why the real-time face recognition can work well by using this equation.
Is there something like principle to explain it?
Looking forward to your reply. :)
The text was updated successfully, but these errors were encountered: