Open CxyZyr opened 1 year ago
I also want to know how to solve this problem, and there are too few datasets. We haven't found any other official data yet
We employed the cycleGAN framework for the generation of near-infrared (NIR) facial images and conducted extensive experiments to demonstrate the effectiveness of this approach. Here are the specific implementation details:
We hope that the methods mentioned above can be of assistance to you.
We employed the cycleGAN framework for the generation of near-infrared (NIR) facial images and conducted extensive experiments to demonstrate the effectiveness of this approach. Here are the specific implementation details:
- We utilized the cycleGAN source code for training.
- We transformed visible light images into grayscale images and trained a cycleGAN model for the transformation from grayscale to NIR images.
- We increased the depth of the generator network(ResNet).
- We pre-trained the above networks on a facial recognition dataset.
We hope that the methods mentioned above can be of assistance to you.
What is the data quality of the IR image generated by vis
in recognition/idmmd part,I seem to have not found the code for generating near-infrared faces from visible light faces. It is unclear from the paper how the mentioned method is implemented. It would be helpful to have detailed implementation steps or code.