activity
20242026
collaborators

5 papers

cs.CV2026

Edge-preserving noise for diffusion models

Jente Vandersanden, Sascha Holl, Xingchang Huang +1

Classical diffusion models typically rely on isotropic Gaussian noise, treating all regions uniformly and overlooking structural information important for high-quality generation.…

cs.CV2025

Restereo: Diffusion stereo video generation and restoration

Xingchang Huang, Ashish Kumar Singh, Florian Dubost +6

Stereo video generation has been gaining increasing attention with recent advancements in video diffusion models. However, most existing methods focus on generating 3D stereoscopic…

cs.LG2025

Multiple Importance Sampling for Stochastic Gradient Estimation

Corentin Salaün, Xingchang Huang, Iliyan Georgiev +2

We introduce a theoretical and practical framework for efficient importance sampling of mini-batch samples for gradient estimation from single and multiple probability distribution…

cs.LG2025

Online Importance Sampling for Stochastic Gradient Optimization

Corentin Salaün, Xingchang Huang, Iliyan Georgiev +2

Machine learning optimization often depends on stochastic gradient descent, where the precision of gradient estimation is vital for model performance. Gradients are calculated from…

cs.CV2024

Blue noise for diffusion models

Xingchang Huang, Corentin Salaün, Cristina Vasconcelos +3

Most of the existing diffusion models use Gaussian noise for training and sampling across all time steps, which may not optimally account for the frequency contents reconstructed b…