A Statistical Benchmark for Diffusion-Posterior-Sampling Algorithms
A walk-through of our benchmark for diffusion-posterior-sampling algorithms, covering inverse problems, the Bayesian framework, and the evaluation pitfalls our work resolves. Check out the interactive visualizations!
On the generalization abilities of diffusion models
Diffusion models as foundation models for inverse problems in imaging (?), and an interesting observation: Vastly different architectures yield similar samples given the same initial conditions.
Denoising with PoGMDMs
A quick illustration of denoising with product of Gaussian mixture diffusion models
A Historical Perspective on Regularizers in Imaging
A quick summary on the historical development of regularizers for inverse problems in imaging, from quadratic gradient penalization over sparsity to deep neural regularizes.
We are back!
Martin Zach is happily relaunching his website. Check it out!