Research map / Learning theory and algorithms / Matrix and tensor methods
Learning Factorized Weight Matrix for Joint Filtering
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 learned sparse residual matrix S (networks eta and kappa) on top of low-rank correlation L (ablation) | factorized weight matrix filter · NYUv2 depth upsampling 4x/8x/16x | RMSE: 2.50 -> 2.16 (4x), 4.57 -> 4.32 (8x), 7.99 -> 7.66 (16x) | Helped |
| add learned sampling locations kappa for the sparse residual (ablation) | factorized weight matrix filter · NYUv2 depth upsampling 4x/8x/16x | RMSE: 2.21 -> 2.16 (4x), 4.43 -> 4.32 (8x), 7.81 -> 7.66 (16x) | Helped |
| add eta without kappa (fixed sampling) over low-rank only (ablation) | factorized weight matrix filter · NYUv2 depth upsampling 4x/8x/16x | RMSE: 2.50 -> 2.21 (4x), 4.57 -> 4.43 (8x), 7.99 -> 7.81 (16x) | Helped |
| use proposed method instead of PAC | depth upsampling · NYUv2 4x | RMSE: 2.39 -> 2.16 | Helped |
| use proposed method instead of PAC | depth upsampling · NYUv2 16x | RMSE: 8.09 -> 7.66 | Helped |
| use proposed method instead of PAC | optical flow upsampling · Sintel 4x | EPE: 0.105 -> 0.096 | Helped |
| use proposed method instead of PAC | optical flow upsampling · Sintel 16x | EPE: 0.592 -> 0.548 | Helped |
| use proposed method instead of SV | depth denoising · Lu et al. benchmark noise 20 | PSNR: 36.44 -> 37.74 | Helped |
| use proposed method instead of SV | depth denoising · Lu et al. benchmark noise 20 | RMSE: 4.02 -> 3.43 | Helped |
| use proposed method instead of NLRN | natural image denoising · Set12 noise 50 | PSNR: 27.64 -> 27.83 | Helped |
| use proposed method instead of NLRN | natural image denoising · BSD68 noise 15 | PSNR: 31.88 -> 31.89 | No clear effect |
| use proposed method instead of NLRN | natural image denoising · Set12 noise 15 | PSNR: 33.16 -> 33.18 | No clear effect |
Closest papers by meaning
- Learning Sparse High Dimensional Filters: Image Filtering, Dense CRFs and Bilateral Neural Networks (CVPR 2016 · similarity 0.85)
- Spatially Variant Linear Representation Models for Joint Filtering (CVPR 2019 · similarity 0.85)
- Deep Weight Factorization: Sparse Learning Through the Lens of Artificial Symmetries (ICLR 2025 · similarity 0.84)
- MaxQ: Multi-Axis Query for N:M Sparsity Network (CVPR 2024 · similarity 0.82)
- Two Sparse Matrices are Better than One: Sparsifying Neural Networks with Double Sparse Factorization (ICLR 2025 · similarity 0.82)
- LERE: Learning-Based Low-Rank Matrix Recovery with Rank Estimation (AAAI 2024 · similarity 0.82)
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.