Research map / Video / Video segmentation and tracking
MetaSCI: Scalable and Adaptive Reconstruction for Video Compressive Sensing
Zhengjue Wang, Hao Zhang, Ziheng Cheng, Bo Chen, Xin Yuan
Check 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 MetaSCI instead of PnP-FFDNet for large-scale reconstruction | SCI reconstruction · 2048×2048×8 large-scale simulation (Average) | PSNR (dB): +2.62dB | Improved |
| Use MetaSCI instead of GAP-TV for large-scale reconstruction | SCI reconstruction · 2048×2048×8 large-scale simulation (Average) | PSNR (dB): 5.13dB | Improved |
| Use MetaSCI rather than BIRNAT for standard-mask reconstruction | 256×256×8 SCI benchmarks · Average | test time per measurement: 0.16 -> 0.025 | Improved |
| Adapt MetaSCI rather than fine-tune BIRNAT for new masks | 256×256×8 SCI benchmarks · Average | adaptation time: 20376 -> 1004 | Improved |
| Adapt MetaSCI rather than fine-tune U-net for new masks | 256×256×8 SCI benchmarks · Average | PSNR (dB): 29.81 -> 31.70 | Improved |
| Adapt the network to the new masks instead of using the original-mask model unchanged | U-net and BIRNAT · 256×256×8 SCI benchmarks, new masks | PSNR (dB): 21.12 -> 29.81; 23.22 -> 33.23 | Improved |
| Increase overlapping size from 0 to 128 (ablation) | MetaSCI large-scale reconstruction · 2048×2048×8 large-scale simulation | PSNR: 31.87 -> 32.49 | Improved |
| Increase overlapping size from 128 to 192 (ablation) | MetaSCI large-scale reconstruction · 2048×2048×8 large-scale simulation | PSNR: 32.49 -> 32.51 | No clear effect |
| Use rank-one kernel modulation instead of full-matrix modulation (ablation) | MetaSCI · SCI model parameterization | meta-modulation parameter count: 537.8k -> 8.7k | Improved |
| Use MetaSCI instead of PnP-FFDNet for large-scale reconstruction | SCI reconstruction · 512×512×8 large-scale simulation (Average) | test time per measurement: 14.22 -> 0.12 | Improved |
Later papers that compared with this method
2 later accepted papers report MetaSCI as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Unfolding Framework with Prior of Convolution-Transformer Mixture and Uncertainty Estimation for Video Snapshot Compressive Imaging (ICCV 2023 · better on 3 of 4 shared results)
- Ensemble Learning Priors Driven Deep Unfolding for Scalable Video Snapshot Compressive Imaging (ECCV 2022 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- BIRNAT — BIRNAT: Bidirectional Recurrent Neural Networks with Adversarial Training for Video Snapshot Compressive Imaging (ECCV 2020) · better on 1 of 2 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
- EfficientSCI: Densely Connected Network With Space-Time Factorization for Large-Scale Video Snapshot Compressive Imaging (CVPR 2023 · similarity 0.92)
- Memory-Efficient Network for Large-Scale Video Compressive Sensing (CVPR 2021 · similarity 0.91)
- Deep Optics for Video Snapshot Compressive Imaging (ICCV 2023 · similarity 0.90)
- DetectSCI: Toward Object-Guided ROI Reconstruction for High-Resolution Video Snapshot Compressive Imaging (CVPR 2026 · similarity 0.90)
- Tensor FISTA-Net for Real-Time Snapshot Compressive Imaging (AAAI 2020 · similarity 0.89)
- Untrained Neural Nets for Snapshot Compressive Imaging: Theory and Algorithms (NeurIPS 2024 · 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.
