Research map / Image generation / Autoregressive and latent generation
RDF-MIG: A Robust Diffusion Framework for Masked Image Generation to Augment Semantic Segmentation and Change Detection
Zian Cao, Wei Wei, Qingshan Gao, Yuanyuan 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 |
|---|---|---|---|
| train diffusion with MCRD loss instead of MSE (ablation) | RDF-MIG · Hi-CNA (SS) image quality | FID: 35.23 -> 27.38 | Improved |
| train diffusion with MCRD loss instead of MSE (ablation) | RDF-MIG · WHU Building image quality | FID: 47.16 -> 17.05 | Improved |
| train diffusion with MCRD loss instead of MSE (ablation) | RDF-MIG · Hi-CNA (CD) image quality | FID: 36.01 -> 29.85 | Improved |
| train diffusion with MCRD loss instead of MSE (ablation) | RDF-MIG synthetic data · Hi-CNA (SS), U-Net | IoU: 42.28 -> 44.13 | Improved |
| train diffusion with MCRD loss instead of MSE (ablation) | RDF-MIG synthetic data · Hi-CNA (CD), SNU-Net | IoU: 45.62 -> 50.00 | Improved |
| use Huber loss instead of MSE (ablation) | RDF-MIG synthetic data · Hi-CNA (SS), U-Net | IoU: 42.28 -> 42.97 | Improved |
| use MCRD instead of Huber loss (ablation) | RDF-MIG synthetic data · Hi-CNA (CD), SNU-Net | IoU: 46.31 -> 50.00 | Improved |
| 15% label noise with MSE loss (ablation) | RDF-MIG synthetic data · Hi-CNA (SS), U-Net | IoU: -2.51 (39.77) | Hurt |
| 15% label noise with MCRD loss (ablation) | RDF-MIG synthetic data · Hi-CNA (SS), U-Net | IoU: -0.28 (43.85) | No clear effect |
| add NIR band generation (multispectral) (ablation) | RDF-MIG · Hi-CNA (SS), U-Net | IoU: 44.13 -> 45.20 | Improved |
| add NIR band generation (multispectral) (ablation) | RDF-MIG · Hi-CNA (CD), SNU-Net | IoU: 50.00 -> 51.07 | Improved |
| add synthetic data to real training set at increasing proportions (ablation) | U-Net / SNU-Net · Hi-CNA and WHU Building | IoU | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ChangeDiff — ChangeDiff: A Multi-Temporal Change Detection Data Generator with Flexible Text Prompts via Diffusion Model (AAAI 2025) · better on 5 of 5 shared results
- SegDiff — SegDiff: Segmented Trajectory Diffusion for Consistent and Adaptive Robot Manipulation (ECCV 2026) · 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
- SatSynth: Augmenting Image-Mask Pairs through Diffusion Models for Aerial Semantic Segmentation (CVPR 2024 · similarity 0.89)
- Task-Oriented Data Synthesis and Control-Rectify Sampling for Remote Sensing Semantic Segmentation (CVPR 2026 · similarity 0.87)
- Changen2: Multi-Temporal Remote Sensing Generative Change Foundation Model (TPAMI 2024 · similarity 0.87)
- A Gift From the Integration of Discriminative and Diffusion-Based Generative Learning: Boundary Refinement Remote Sensing Semantic Segmentation (TPAMI 2026 · similarity 0.87)
- DiffusionSat: A Generative Foundation Model for Satellite Imagery (ICLR 2024 · similarity 0.86)
- The Change You Want To Detect: Semantic Change Detection In Earth Observation With Hybrid Data Generationf (CVPR 2025 · 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.
