Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Variational Inference for Computational Imaging Inverse Problems
Francesco Tonolini, Jack Radford, Alex Turpin, Daniele Faccio, Roderick Murray-Smith
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train with proposed multi-fidelity variational framework instead of paired examples only | CVAE · CelebA K=1,000 x2 down-sampling | test ELBO: -19827 -> 14553 | Improved |
| train with proposed framework instead of paired examples only | CVAE · CelebA K=10,000 x2 down-sampling | test ELBO: 13450 -> 14763 | Improved |
| train with proposed framework instead of simulations only | CVAE · CelebA K=1,000 x2 down-sampling | test ELBO: -208 -> 14553 | Improved |
| train with proposed framework instead of naive paired+simulations pooling | CVAE · CelebA K=1,000 x2 down-sampling | test ELBO: 11300 -> 14553 | Improved |
| train with proposed framework instead of naive pooling | CVAE · CIFAR10 K=10,000 x2 down-sampling | test ELBO: 12096 -> 14415 | Improved |
| increase paired examples K from 1,000 to 10,000 for standard CVAE (ablation) | CVAE paired only · CelebA blurring sigma=1.5px | test ELBO: -13872 -> 13502 | Improved |
| increase paired examples K from 1,000 to 10,000 for proposed framework (ablation) | proposed framework · CelebA partial occlusion | test ELBO: 15134 -> 15187 | No clear effect |
| add unpaired targets and low-fidelity model (proposed) for scarce paired data | CVAE · CelebA 64x64 deblurring, K<few thousand | PSNR and ELBO | Improved |
| use more accurate observation model (ablation) | proposed framework · CelebA blurring, low K | PSNR | Improved |
| use proposed variational framework instead of HIO and deep ANN | holographic reconstruction · MNIST Fourier intensity (9,600 pairs) | reconstruction quality | Improved |
Closest papers by meaning
- Principled Probabilistic Imaging using Diffusion Models as Plug-and-Play Priors (NeurIPS 2024 · similarity 0.86)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.85)
- Sparse Bayesian Deep Learning for Cross Domain Medical Image Reconstruction (AAAI 2024 · similarity 0.85)
- How many measurements are enough? Bayesian recovery in inverse problems with general distributions (NeurIPS 2025 · similarity 0.85)
- Improving Diffusion Models for Inverse Problems Using Optimal Posterior Covariance (ICML 2024 · similarity 0.84)
- Solving Inverse Problems with FLAIR (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
