Search Results for author:
Found 4 papers, 3 papers with code
OMG: Occlusion-friendly Personalized Multi-concept Generation in Diffusion Models
We also observe that the initiation denoising timestep for noise blending is the key to identity preservation and layout.
Dual Teacher Knowledge Distillation with Domain Alignment for Face Anti-spoofing
In this paper, we propose a domain adversarial attack (DAA) method to mitigate the training instability problem by adding perturbations to the input images, which makes them indistinguishable across domains and enables domain alignment.
Fingerprint Presentation Attack Detection by Channel-wise Feature Denoising
The proposed method learns important features of fingerprint images by weighing the importance of each channel and identifying discriminative channels and "noise" channels.
Taming Self-Supervised Learning for Presentation Attack Detection: De-Folding and De-Mixing
Even though there are numerous Presentation Attack Detection (PAD) techniques based on both deep learning and hand-crafted features, the generalization of PAD for unknown PAI is still a challenging problem.
Cannot find the paper you are looking for? You can
Submit a new open access paper.
