search

Shawn-Shan / fawkes

Fawkes, privacy preserving tool against facial recognition systems. More info at https://sandlab.cs.uchicago.edu/fawkes
https://sandlab.cs.uchicago.edu/fawkes
BSD 3-Clause "New" or "Revised" License
5.22k stars 481 forks source link

difference of low , mid , high_extract.h5 #100

Closed ybsu closed 3 years ago

ybsu commented 3 years ago

Dear sir, I am a little confused of these models: mid_extract.h5, high_extract.h5, low_extract.h5，is the performance of low_extract.h5 worse than mid_extract.h5 and high_extract.h5, or high_extract.h5 is used to generate high cloakes, low_extract.h5 is used to generate low cloakes ? Is there any connection between the model's performance and the name of model like "low", "high"? thanks!

ybsu commented 3 years ago

Dear sir, one more question, I have trained some models based on tensorflow, how can I convert tensorflow model to h5 model based on keras that can be used directly in your code, I have no idea about this , grateful to get advice from u,sir! Thank you .

Shawn-Shan commented 3 years ago

The mid low and high_ has no real meaning, - just some names I gave arbitrarily when doing the experiments.

As for converting tf models to .h5 models, it is pretty complicated. I suggest you to change the source code of Fawkes to directly read your tf graph.