Faceres Original Model

[ceyxasm]

Hello @vladmandic
You have given credits to HSE_FaceRes_tf for face description models. Moreover, you provide two models for face description calculation.
One is faceres and other is faceres-deep. I have a few queries

1. What is the difference between the two models: faceres and faceres-deep?
2. Which one of the two corresponds to the original model provided by HSE_FaceRes_tf
3. I generated embeddings using all the three models; 2 using human and 1 from HSE_FaceRes_tf; but no two match. Is this expected? Or am I missing some image preprocessing?

Motivation: I want to run a model in my python backend to store the embeddings of images of people. When same people are captured by human; I am using descriptors given by human to do a recognition test.

To generate human embeddings I used your code and to generate embeddings from HSE_FaceRes_tf, I used their code @ here.
I also ran their code with ImageNet normalizations and RGB->BGR but nothing worked.

[ceyxasm]

Adding further to this; the way currently the model calculates scores is not super discriminative.

Is calculating euclidean distance between the embeddings to calculate the score the best method here??