Python: WEBSERVICE RESTFULL to compare 2 faces and counting faces using sended images.

I have 2 of this libraries searched on google.

Try to make it work i can because some
this lib libnvinfer.so.7 is new and dont have this old lib
avaliable.

https://stackoverflow.com/questions/60368298/could-not-load-dynamic-library-libnvinfer-so-6

https://pypi.org/project/face-compare/

Linux Version is Ubuntu 22.04 with NVIDIA drivers
installed GTX 660.

Another lib:

https://github.com/ageitgey/face_recognition

I need create a webservice fresk / gnunecorn
one route to compare 2 photos and tell-me
if the person on photo is the same
person.

Autentication using barrier token need be implanted.
the key will be on source code of aplication.

ROUTE ONE:

https://address/facecompare METHOD: POST

i will pass 2 images for the service you need download
the images make compare and returns if people match or
not.

REQUEST SEND:

{
image_1: "https://address.domain.ltd/photo1.jpg",
image_2: "https://address.domain.ltd/photo1.jpg",
tolerance: 0.6 // if aplicable acording with used lib
}


{
image_1: "data:image/jpeg;base64,/9j/4AAQSkZJRgABAQEARwBHAAD",
image_2: "data:image/jpeg;base64,/9j/4AAQSkZJRgABAQEARwBHAAD",
tolerance: 0.6 // if aplicable acording with used lib
}

{
image_1: "https://address.domain.ltd/photo1.jpg",
image_2: "data:image/jpeg;base64,/9j/4AAQSkZJRgABAQEARwBHAAD"
tolerance: 0.6 // if aplicable acording with used lib
}

can use another extra parameters acording lib used for it.
important of this project is possible use CPU or GPU for
speed ups compare things.

RETURNS:

{
match: TRUE or FALSE,
percent: 0.0 /match % if aplicable.
}

https://address/facecount METHOD: POST

{
image: "https://address.domain.ltd/photo1.jpg"
}

OR

{
image_1: "data:image/jpeg;base64,/9j/4AAQSkZJRgABAQEARwBHAAD",
}


I cand send image by url or base64 encoded image or a combination
beetween to of this one encoded and one url.

RETURNS:

{
counted: 0 // 0 no faces found and 1 or more how many faces was ben found
}

I need a complete howto how install this solution from scratch and
able to use this API using CPU or GPU and make it configurable.

If you have better library you are welcome to sugest.