Research map / Image generation / Image restoration
Neural Proximal Gradient Descent for Compressive Imaging
Morteza Mardani, Qingyun Sun, David Donoho, Vardan Papyan, Hatef Monajemi, Shreyas Vasanawala, John Pauly
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 |
|---|---|---|---|
| Use recurrent proximal-gradient iterations with a shared small ResNet instead of a single deep ResNet | MRI reconstruction · Pediatric abdominal MRI test | SNR (dB): 24.33 -> 26.07 | Improved |
| Use recurrent proximal-gradient iterations with a shared small ResNet instead of CS-WV | MRI reconstruction · Pediatric abdominal MRI test | SNR (dB): 22.51 -> 26.07 | Improved |
| Use recurrent proximal-gradient iterations with a shared small ResNet instead of CS-WV | MRI reconstruction · Pediatric abdominal MRI test | Inference time (sec): 1.16 -> 0.04 | Improved |
| Use 5 iterations and 2 RBs instead of 10 iterations and 1 RB (ablation) | Neural proximal gradient descent · Pediatric abdominal MRI test | SNR (dB): 26.07 -> 26.94 | Improved |
| Use 5 iterations and 2 RBs instead of 10 iterations and 1 RB (ablation) | Neural proximal gradient descent · Pediatric abdominal MRI test | SSIM: 0.9117 -> 0.9221 | Improved |
| Use 5 iterations and 2 RBs instead of 10 iterations and 1 RB (ablation) | Neural proximal gradient descent · Pediatric abdominal MRI test | Inference time (sec): 0.04 -> 0.10 | Hurt |
| Use 2 iterations and 5 RBs instead of 10 iterations and 1 RB (ablation) | Neural proximal gradient descent · Pediatric abdominal MRI test | SNR (dB): 26.07 -> 26.55 | Improved |
| Use 2 iterations and 5 RBs instead of 10 iterations and 1 RB (ablation) | Neural proximal gradient descent · Pediatric abdominal MRI test | Inference time (sec): 0.04 -> 0.12 | Hurt |
| Add more than 5 RBs to a single-iteration network (ablation) | MRI reconstruction · Pediatric abdominal MRI | SNR (dB): 24.15; 24.33 | No clear effect |
| Increase recurrent copies from one to three (ablation) | Neural proximal gradient descent with 1 RB · Pediatric abdominal MRI | SNR (dB): more than 2dB | Improved |
| Increase recurrent copies from one to three (ablation) | Neural proximal gradient descent with 2 RBs · Pediatric abdominal MRI | SNR (dB): more than 3dB | Improved |
Closest papers by meaning
- Recurrent Variational Network: A Deep Learning Inverse Problem Solver Applied to the Task of Accelerated MRI Reconstruction (CVPR 2022 · similarity 0.89)
- A Unified Model for Compressed Sensing MRI Across Undersampling Patterns (CVPR 2025 · similarity 0.89)
- Learning Proximal Operators: Using Denoising Networks for Regularizing Inverse Imaging Problems (ICCV 2017 · similarity 0.88)
- Compressed Sensing MRI Using a Recursive Dilated Network (AAAI 2018 · similarity 0.87)
- Proximal Algorithm Unrolling: Flexible and Efficient Reconstruction Networks for Single-Pixel Imaging (CVPR 2025 · similarity 0.87)
- What's in a Prior? Learned Proximal Networks for Inverse Problems (ICLR 2024 · similarity 0.87)
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.
