Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-Supervised Neural Networks for Spectral Snapshot Compressive Imaging
Ziyi Meng, Zhenming Yu, Kun Xu, 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 plug-and-play framework around DIP instead of sole DIP (ablation) | PnP-DIP · CASSI synthetic, 10 scenes avg | PSNR (dB): 26.99 -> 30.48 | Improved |
| use double fidelity terms instead of single fidelity term (ablation) | PnP-DIP · CASSI synthetic, 10 scenes avg | PSNR (dB): 28.87 -> 30.48 | Improved |
| use DIP U-net prior instead of deep decoder (ablation) | PnP-DIP · CASSI synthetic, 10 scenes avg | PSNR (dB): 28.44 -> 30.48 | Improved |
| use ConvTranspose upsampling instead of bilinear (ablation) | PnP-DIP · CASSI synthetic, 10 scenes avg | PSNR (dB): 29.69 -> 30.48 | Improved |
| use ConvTranspose upsampling instead of PixelShuffle (ablation) | PnP-DIP · CASSI synthetic, 10 scenes avg | PSNR (dB): 27.59 -> 30.48 | Improved |
| use average-pooling instead of max-pooling (ablation) | PnP-DIP · CASSI synthetic, 10 scenes avg | PSNR (dB): 30.26 -> 30.48 | Improved |
| add pre-trained HSI deep denoiser to PnP-DIP (ablation) | PnP-DIP · CASSI synthetic, 10 scenes avg | PSNR (dB): 30.48 -> 31.30 | Improved |
| compare untrained PnP-DIP with supervised TSA-Net | PnP-DIP · CASSI synthetic, 10 scenes avg | PSNR (dB): 30.15 -> 30.48 | Improved |
| compare untrained PnP-DIP with supervised TSA-Net on SSIM | PnP-DIP · CASSI synthetic, 10 scenes avg | SSIM: 0.893 -> 0.854 | Hurt |
| add Poisson noise SNR=25dB to measurements (supervised TSA-Net) | TSA-Net · CASSI synthetic, 10 scenes avg | PSNR (dB): 30.15 -> 24.15 | Hurt |
| add Poisson noise SNR=25dB to measurements (PnP-DIP) | PnP-DIP · CASSI synthetic, 10 scenes avg | PSNR (dB): 30.48 -> 27.73 | Hurt |
| use a larger 60-pixel real-mask shift instead of 30-pixel binary mask: PnP-DIP vs PnP-HSI gap | PnP-DIP · ICVL | PSNR (dB): 32.91 -> 37.27 | Improved |
Closest papers by meaning
- Untrained Neural Nets for Snapshot Compressive Imaging: Theory and Algorithms (NeurIPS 2024 · similarity 0.93)
- Effective Snapshot Compressive-Spectral Imaging via Deep Denoising and Total Variation Priors (CVPR 2021 · similarity 0.91)
- Self-Supervised One-Step Diffusion Refinement for Snapshot Compressive Imaging (AAAI 2026 · similarity 0.90)
- SAH-SCI: Self-Supervised Adapter for Efficient Hyperspectral Snapshot Compressive Imaging (ECCV 2024 · similarity 0.90)
- Memory-Efficient Network for Large-Scale Video Compressive Sensing (CVPR 2021 · similarity 0.90)
- Self-supervised Bayesian Deep Learning for Image Recovery with Applications to Compressive Sensing (ECCV 2020 · similarity 0.89)
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.
