How can I Integrate A face Detection model with an already fine$-$tuned ConvNeXt image classifier for Face Recognition?

Hello,

I need advice on how to move on with my project, Initially I wanted to create a face recognition system. I first gathered a dataset of celebrity faces with $99$ classes and about $16$k total images and fine$-$tuned ConvNeXtTiny model on the dataset using tensorflow and got a result of $93\%$ accuracy. Now this is technically only an image classification application where it can tell the faces apart and tell which celebrity it is$.$ However, I need to extened this project to a fully face recognition system.

How can I use tensorflow transfer learning with existing models to make this system full circle? Basically I need a face detection model that is compatible with tensorflow $2\mathrm{.}15\mathrm{.}0$ then preprocess the faces$($Either from a webcam or can be processed from an unknown dataset$)$ then passing them to the ConvNeXt model for recognition. my Idea is that the unknown faces would be registered and added to the dataset.

I have done some research and tried to implement VGGFACE but I was met with so many errors that i couldn't go forward with it because apparently VGGface isnt compatible with tensorflow $2.$x $>.$

I need recommendations and guidance on how to move forward and integrate a model with my face image classifier model. are there any resources that can be implemented easily with tensorflow $?$ And how easy or hard is this task to complete?