I have 2 questions about the pre-trained models.
Which datasets the models were pre-trained on?
How do they evaluate their performance on LFW? do they fine-tune or they change the number of dense neurons?

Thanks for your efforts, your work is really helpful.

Please add this model to next release so the YOLO model's path can be updated in deepface.
Referring to this issue serengil/deepface#939

yolov8n_face.pt.zip

How to calculate the accuracy of the model in these data sets(.h5) file?
Please advise

Hi,

https://github.com/davidsandberg/facenet/blob/master/README.md listed two pretrained models:

Pre-trained models

Obviously the 20180402-114759 has a higher LFW accuracy. I would like to know which model was facenet512_weights.h5 converted from?

first off , thanks for your great work,
secondly, could you help us who blocked by GFW, very slow download speed.
to zip all of pre-trained model files as one?
it will be very helpful for us download it.
thanks in advance.

First of all, thank you for the great job you are doing. This lib is amazing and it's helping me a lot with my BSc thesis (and I'll be referencing it appropriately).

However, there is an essential piece of information that I couldn't find anywhere, not even in your blog posts (maybe I didn't search properly): Which datasets were the models pre-trained on?

I saw that someone asked the same question on issue #2, but I don't think it was actually answered.