Research map / Image generation / Image restoration
Self-supervised Bayesian Deep Learning for Image Recovery with Applications to Compressive Sensing
Tongyao Pang, Yuhui Quan, Hui Ji
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 |
|---|---|---|---|
| add random Bayesian weights (vs MLE deterministic weights) (ablation) | Self-supervised BNN · Set11 CS (noiseless, ratio 25%) | PSNR (dB): 29.43 -> 32.30 | Improved |
| add random Bayesian weights (vs MAP deterministic weights) (ablation) | Self-supervised BNN · Set11 CS (noiseless, ratio 25%) | PSNR (dB): 29.51 -> 32.30 | Improved |
| add random Bayesian weights (vs MLE deterministic weights) (ablation) | Self-supervised BNN · Set11 CS (noise 10, ratio 25%) | PSNR (dB): 27.24 -> 28.67 | Improved |
| add random Bayesian weights (vs MAP deterministic weights) (ablation) | Self-supervised BNN · Set11 CS (noise 10, ratio 4%) | PSNR (dB): 20.25 -> 21.91 | Improved |
| add random Bayesian weights (vs MLE deterministic weights) (ablation) | Self-supervised BNN · Set11 CS (noiseless, ratio 4%) | PSNR (dB): 21.13 -> 23.26 | Improved |
| add Gaussian weight prior (MAP vs MLE) to deterministic network (ablation) | deterministic network · Set11 CS (noiseless, ratio 4%) | PSNR (dB): 21.13 -> 21.08 | No clear effect |
| add Gaussian weight prior (MAP vs MLE) to deterministic network (ablation) | deterministic network · Set11 CS (noise 10, ratio 25%) | PSNR (dB): 27.24 -> 27.18 | No clear effect |
| use BNN instead of DIP for CS reconstruction | Self-supervised BNN · Set11 CS (noise 10, ratio 40%) | PSNR (dB): 28.87 -> 30.39 | Improved |
| use BNN instead of DIP for CS reconstruction | Self-supervised BNN · MRI 2D Gaussian mask (noise 10%) | PSNR (dB): 24.41 -> 27.12 | Improved |
| use BNN instead of DIP for CS reconstruction | Self-supervised BNN · MRI 1D Gaussian mask (noiseless) | PSNR (dB): 31.80 -> 31.38 | Hurt |
| use self-supervised BNN instead of supervised ISTA-Net | Self-supervised BNN · Set11 CS (noiseless, ratio 40%) | PSNR (dB): 35.97 -> 35.71 | Hurt |
| use self-supervised BNN instead of supervised ISTA-Net | Self-supervised BNN · Set11 CS (noise 10, ratio 40%) | PSNR (dB): 27.98 -> 30.39 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ISTA-Net — ISTA-Net: Interpretable Optimization-Inspired Deep Network for Image Compressive Sensing (CVPR 2018) · better on 7 of 10 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Self-Supervised Deep Image Restoration via Adaptive Stochastic Gradient Langevin Dynamics (CVPR 2022 · similarity 0.90)
- Dual-Domain Self-Supervised Learning and Model Adaption for Deep Compressive Imaging (ECCV 2022 · similarity 0.89)
- Self-Supervised Neural Networks for Spectral Snapshot Compressive Imaging (ICCV 2021 · similarity 0.89)
- Training Deep Learning Based Image Denoisers From Undersampled Measurements Without Ground Truth and Without Image Prior (CVPR 2019 · similarity 0.88)
- Untrained Neural Nets for Snapshot Compressive Imaging: Theory and Algorithms (NeurIPS 2024 · similarity 0.86)
- Sparse Bayesian Deep Learning for Cross Domain Medical Image Reconstruction (AAAI 2024 · similarity 0.85)
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.
