Research map / Video / Video segmentation and tracking
Hierarchical Separable Video Transformer for Snapshot Compressive Imaging
Ping Wang, Yulun Zhang, Lishun Wang, 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 |
|---|---|---|---|
| use HiSViT_9 instead of CTM-SCI | video SCI reconstruction · Grayscale SCI benchmark | PSNR (dB): 36.52 -> 37.00 with 12.00% of MACs | Improved |
| use HiSViT_9 instead of EfficientSCI | video SCI reconstruction · Color SCI benchmark | PSNR (dB): 37.49 -> 38.01 | Improved |
| use frame-wise 2D feature extraction instead of 3D CNN (disable early temporal interaction) (ablation) | HiSViT_9 · Grayscale SCI benchmark | PSNR (dB): 36.65 -> 36.79 | Improved |
| use RSTB instead of 3D CNN for reconstruction (ablation) | HiSViT_9 · Grayscale SCI benchmark | PSNR (dB): 36.86 -> 36.92 | Improved |
| add skip connection between shallow and upsampled refined features (ablation) | HiSViT_9 · Grayscale SCI benchmark | PSNR (dB): 36.92 -> 37.00 | Improved |
| add separable (spatial-then-temporal) design to CSS-MSA (ablation) | CSS-MSA · Grayscale SCI benchmark | PSNR (dB): 36.57 -> 37.00; MACs 2155.08 -> 1535.92 | Improved |
| add cross-scale design to CSS-MSA (ablation) | CSS-MSA · Grayscale SCI benchmark | PSNR (dB): 36.92 -> 37.00; MACs 1769.78 -> 1535.92 | Improved |
| use CSS-MSA instead of SW-MSA or FW-MSA (ablation) | HiSViT_9 · Grayscale SCI benchmark | PSNR (dB): 35.55 (SW-MSA), 35.38 (FW-MSA) -> 37.00 | Improved |
| use GSM-FFN instead of regular FFN (ablation) | HiSViT_9 · Grayscale SCI benchmark | PSNR (dB): 36.28 -> 37.00 | Improved |
| add gated self-modulation and factorized STConv to FFN (ablation) | GSM-FFN · Grayscale SCI benchmark | PSNR (dB): 36.82 (w/o GSM), 36.86 (w/o STConv) -> 37.00 | Improved |
| scale HiSViT from 9 to 13 blocks | HiSViT · Grayscale SCI benchmark | PSNR (dB): 37.00 -> 37.29 | 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.
- 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
- Spatial-Temporal Transformer for Video Snapshot Compressive Imaging (TPAMI 2022 · similarity 0.91)
- EfficientSCI: Densely Connected Network With Space-Time Factorization for Large-Scale Video Snapshot Compressive Imaging (CVPR 2023 · similarity 0.90)
- Unfolding Framework with Prior of Convolution-Transformer Mixture and Uncertainty Estimation for Video Snapshot Compressive Imaging (ICCV 2023 · similarity 0.90)
- Memory-Efficient Network for Large-Scale Video Compressive Sensing (CVPR 2021 · similarity 0.88)
- MetaSCI: Scalable and Adaptive Reconstruction for Video Compressive Sensing (CVPR 2021 · similarity 0.87)
- Untrained Neural Nets for Snapshot Compressive Imaging: Theory and Algorithms (NeurIPS 2024 · similarity 0.86)
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.
