Open ChienLady opened 2 years ago
The formula is (tanh ((coef0 - distance) * coef1) + 1) / 2
coef0 is an average between max and min possible euclidian distances (without outliners).
coef1 tries to put negative scenarios to 0 and positive ones to 1
In reality, it's lots of experiments (we did 200 000) and just a formula
If you provide me with a model, I can reproduce those experiments. What is the license of the model?
Thank you for your explanation,
This is the google drive link of this model: https://drive.google.com/drive/folders/11TI4Gs_lO-fbts7cgWNqvVfm9nps2msE?usp=sharing
More information in https://github.com/ZhaoJ9014/face.evoLVe#model-zoo The model i am using is trained on the Private Asia Face dataset.
Thanks.
The repository you send me uses Pytorch. And our InsightFace implementation is MXNet based. So CompreFace can't read Pytorch models.
Can you share me the script you calculated the coef, i want to learn by code, i can't find it in CompareFace repo My gmail: chientranminh0511@gmail.com
Thank you.
Hi, would you mind sharing the script to me as well.
Hi, this is the awesome face application, i have a question
How can i find the status (coef0, coef1) of my arcface model ? I use Asia-ArcFace model of Face Evolved
Thank you