Research map / Video / Video segmentation and tracking
Unfolding Framework with Prior of Convolution-Transformer Mixture and Uncertainty Estimation for Video Snapshot Compressive Imaging
Siming Zheng, Xin Yuan
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 uncertainty-map features as regularized prior in each phase (ablation) | CTM unfolding network · Video SCI benchmark | PSNR (dB): 36.29 -> 36.52 | Improved |
| add 3D blocked dense attention (BDA) (ablation) | CTM · Video SCI benchmark | PSNR (dB): 31.11 -> 36.29 | Improved |
| add 3D dilated sparse attention (DSA) (ablation) | CTM · Video SCI benchmark | PSNR (dB): 31.04 -> 36.29 | Improved |
| add 3D-CNN feature fusion (FF) (ablation) | CTM · Video SCI benchmark | PSNR (dB): 27.81 -> 36.29 | Improved |
| replace DSA with BDA at same parameters and FLOPs (ablation) | CTM · Video SCI benchmark | PSNR (dB): -0.33 dB | Hurt |
| replace BDA with DSA at same parameters and FLOPs (ablation) | CTM · Video SCI benchmark | PSNR (dB): -0.87 dB | Hurt |
| change the order of BDA, DSA and FF sub-modules (ablation) | CTM block · Video SCI benchmark | PSNR (dB) | No clear effect |
| train reconstruction directly with the uncertainty loss L_U (ablation) | CTM unfolding network · Video SCI benchmark | PSNR (dB): -1.1 dB | Hurt |
| increase unfolding phases from three to four (ablation) | CTM unfolding network · Video SCI benchmark | PSNR (dB): 36.29 -> 36.37 | No clear effect |
| increase unfolding phases from one to three (ablation) | CTM unfolding network · Video SCI benchmark | PSNR (dB): 35.51 -> 36.29 | Improved |
| estimate uncertainty map with a one-phase network rather than three-phase (ablation) | uncertainty estimation · Video SCI benchmark | PSNR (dB) | No clear effect |
| evaluate with unseen masks without retraining | Ours-with-Uncertainty · Video SCI benchmark | PSNR (dB): 36.52 -> 36.47 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MetaSCI — MetaSCI: Scalable and Adaptive Reconstruction for Video Compressive Sensing (CVPR 2021) · better on 3 of 4 shared results
- BIRNAT — BIRNAT: Bidirectional Recurrent Neural Networks with Adversarial Training for Video Snapshot Compressive Imaging (ECCV 2020) · better on 1 of 1 shared result
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
- Dense Deep Unfolding Network With 3D-CNN Prior for Snapshot Compressive Imaging (ICCV 2021 · similarity 0.91)
- Memory-Efficient Network for Large-Scale Video Compressive Sensing (CVPR 2021 · similarity 0.91)
- Ensemble Learning Priors Driven Deep Unfolding for Scalable Video Snapshot Compressive Imaging (ECCV 2022 · similarity 0.91)
- Untrained Neural Nets for Snapshot Compressive Imaging: Theory and Algorithms (NeurIPS 2024 · similarity 0.90)
- Dual Prior Unfolding for Snapshot Compressive Imaging (CVPR 2024 · similarity 0.89)
- Using Powerful Prior Knowledge of Diffusion Model in Deep Unfolding Networks for Image Compressive Sensing (CVPR 2025 · similarity 0.88)
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.
