CGCL-codes/AMT-GAN
The official implementation of our CVPR 2022 paper "Protecting Facial Privacy: Generating Adversarial Identity Masks via Style-robust Makeup Transfer".
PythonMIT
Issues
- 1
CUDA device problem
#21 opened by Leo1214 - 2
We do not know the matchup of the `(source, target)` for evaluation(i.e. we do not know the matchup of the images in different groups and the target identity that you have aimed) although you have uploaded `pair.zip` and `assets.zip`.
#16 opened by yuuma002 - 0
- 0
How to set the threshold on different datasets to evaluate the FR models using TAR@FAR？
#24 opened by AaronSun2000 - 0
How to convert to ONNX
#23 opened by ZTMIDGO - 0
Why
#22 opened by 1215372621 - 0
- 0
LPIPS loss
#19 opened by Werty18 - 1
attack success rate seems different from paper
#18 opened by yuxu915 - 1
About multi-GPU training models
#17 opened by godHhh - 1
Could you provide the image `pairs` you have used for reproducing the result in Table.1 of the paper?
#15 opened by yuuma002 - 3
Have the MI-FGSM, TI-DIM, and TIP-IM in the paper used the ensemble attack strategy, and are the maximum perturbation of PGD, MI-FGSM, TI-DIM, and TIP-IM all 16/255 under the L Infinity distance?
#14 opened by GeorgeCarpenter - 3
- 1
- 7
Hello, does the PGD attack in this paper use the integrated attack strategy
#12 opened by SuperCarryFu - 3
- 1
the question of training epoches
#11 opened by Yueming6568 - 1
What's the `gamma` value of TIP-IM you have set for evaluating the performance of TIP-IM?
#10 opened by GeorgeCarpenter - 1
- 2
How to change `IMG_SIZE`?
#8 opened by samuelyutt - 3
Can we feed the facenet model in the repo with the image whose size is (112, 112)?
#7 opened by GeorgeCarpenter - 1
Does the trained AMT-GAN generator can only attack a single target identity?
#5 opened by GeorgeCarpenter - 2
Cannot install skimage==0.0
#3 opened by samuelyutt - 2
- 2
Could you release your dataset and pre-trained models for reproducing your work?
#1 opened by zhoufengfan