Research map / Image generation / Super-resolution
Learning to Balance: Decoupled Siamese Diffusion Transformer for Reference-Based Remote Sensing Image Super-Resolution
Bin Luo, Runmin Dong, Zhaoyang Luo, Jinxiao Zhang, Jiyao Zhao, Fan Wei, Haohuan Fu
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 |
|---|---|---|---|
| replace unconstrained M3-Attention with decoupled siamese attention (ablation) | M3-DiT · SECOND x8 | FID: 22.66 -> 19.16 | Improved |
| add Patch-Level Weighting module (ablation) | Siamese attention baseline · SECOND x8 | LPIPS: 0.2362 -> 0.2291 | Improved |
| add Patch-Level Weighting module (ablation) | Siamese attention baseline · SECOND x16 | FID: 31.29 -> 30.32 | Improved |
| add autoguidance (ablation) | Siamese attention baseline · SECOND x8 | FID: 19.16 -> 18.56 | Improved |
| add autoguidance (ablation) | Siamese attention baseline · SECOND x16 | FID: 31.29 -> 27.96 | Improved |
| add both PLW and autoguidance (ablation) | Siamese attention baseline · SECOND x8 | FID: 19.16 -> 15.87 | Improved |
| add both PLW and autoguidance (ablation) | Siamese attention baseline · SECOND x16 | FID: 31.29 -> 21.58 | Improved |
| inject LR and Ref via independent ZeroLinear (Variant A) instead of PLW (ablation) | Siamese attention baseline · SECOND x8 | FID: 19.16 -> 23.07 | Hurt |
| inject summed LR and Ref via one ZeroLinear (Variant B) instead of PLW (ablation) | Siamese attention baseline · SECOND x8 | FID: 19.16 -> 19.81 | No clear effect |
| use PLW injection (Variant C) instead of Variant A (ablation) | Siamese attention baseline · SECOND x8 | FID: 23.07 -> 18.93 | Improved |
| set guidance coefficient omega to 0 instead of 1.0 (ablation) | DS-DiT autoguidance · SECOND x8 | FID: 18.93 -> 75.14 | Hurt |
| raise guidance coefficient omega from 1.2 to 1.5 (ablation) | DS-DiT autoguidance · SECOND x8 | FID: 15.87 -> 22.14 | Hurt |
Closest papers by meaning
- Building Bridges across Spatial and Temporal Resolutions: Reference-Based Super-Resolution via Change Priors and Conditional Diffusion Model (CVPR 2024 · similarity 0.93)
- Remote Sensing Image Super-Resolution for Imbalanced Textures: A Texture-Aware Diffusion Framework (CVPR 2026 · similarity 0.91)
- BUFF: Bayesian Uncertainty Guided Diffusion Probabilistic Model for Single Image Super-Resolution (AAAI 2025 · similarity 0.89)
- SinSR: Diffusion-Based Image Super-Resolution in a Single Step (CVPR 2024 · similarity 0.88)
- When Latents Forget Pixels: Restoring Fidelity in Diffusion Transformer Super-Resolution (NeurIPS 2026 · similarity 0.88)
- Pixel-level and Semantic-level Adjustable Super-resolution: A Dual-LoRA Approach (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.
