Large-scale identity-labeled face datasets are essential for training modern face recognition models, yet collecting and annotating such data is costly and raises privacy and ethical concerns. Existing synthetic face generation methods typically derive identities from predefined class labels, unconditional image genera...
F. Boutros, Meric Demirors, Darian Tomašević et al.· 0 citations
Recent diffusion-based image generation models have enabled new directions in virtual try-on and garment synthesis. However, existing approaches often trade off between controllability and accessibility: tuning-based methods typically require additional training or specialized inputs, while tuning-free methods may lack...
Ajda Lampe, Ana Cvetković, Dong Seog Han et al.· International Symposium ELMA...· 0 citations
A novel (robust) face de-identification approach that combines target-oriented models with a dedicated diffusion process specifically designed to destroy any imperceptible information that may allow learning to reverse the de-identification procedure, and exhibits state-of-the-art fidelity and attribute-retention capab...
Felix Rosberg, Vitomir Štruc, Cristofer Englund et al.· arXiv.org· 0 citations
This work proposes GRACE, a structured concept erasure framework designed to enable localized and selective intervention, and introduces a semantically weighted sensitive subspace estimation to precisely lock intervention directions, and employs lightweight subspace-constrained adapters to prevent global semantic distu...
Experiments with face recognition models using CNN, ViT, and Swin backbones show that SCOUT discovers interpretable concepts beyond standard attribute labels and enables controllable, identity-aware template manipulation with negligible impact on identity matching.
Leon Todorov, Peter Rot, Peter Peer et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.