Research map / Image generation / Fast diffusion sampling
Pattern-Structure Diffusion for Multi-Task Learning
Check 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 |
|---|---|---|---|
| Use PSD for multi-task segmentation | PSD-ResNet50 · NYUD-v2 | mIoU: 48.4 -> 51.0 | Helped |
| Use PSD for multi-task segmentation | PSD-ResNet50 · SUNRGB-D | mIoU: 50.5 -> 50.6 | Helped |
| Use joint depth and surface-normal PSD with 12k training images | PSD-Td+Tn · NYUD-v2 | RMSE: 0.501 -> 0.488 | Helped |
| Use PSD for surface-normal prediction | PSD-ResNet50 · NYUD-v2 | Mean: 19.0 -> 18.2 | Helped |
| Add intra-task PSD alone (ablation) | Joint three-task baseline · NYUD-v2 | mIoU: 40.9 -> 42.2 | Helped |
| Add intra-task PSD alone (ablation) | Joint three-task baseline · NYUD-v2 | RMSE: 0.585 -> 0.578 | Hurt |
| Add inter-task PSD alone (ablation) | Joint three-task baseline · NYUD-v2 | mIoU: 40.9 -> 43.1 | Helped |
| Add both intra-task and inter-task PSD at all scales (ablation) | Joint three-task baseline · NYUD-v2 | mIoU: 40.9 -> 50.0 | Helped |
| Apply both PSD modules at all scales rather than the small scale only (ablation) | Joint three-task baseline · NYUD-v2 | RMSE: 0.548 -> 0.498 | Helped |
| Jointly train segmentation, depth, and surface normal rather than each task separately (ablation) | PSD · NYUD-v2 | mIoU: 42.4 -> 44.9 | Helped |
| Jointly train segmentation, depth, and surface normal rather than each task separately (ablation) | PSD · NYUD-v2 | RMSE: 0.572 -> 0.548 | Helped |
Closest papers by meaning
- DiffusionMTL: Learning Multi-Task Denoising Diffusion Model from Partially Annotated Data (CVPR 2024 · similarity 0.87)
- Denoising Task Routing for Diffusion Models (ICLR 2024 · similarity 0.86)
- Towards Principled Task Grouping for Multi-Task Learning (NeurIPS 2026 · similarity 0.84)
- Diffusion Models for Multi-Task Generative Modeling (ICLR 2024 · similarity 0.84)
- Towards Universal Dataset Distillation via Task-Driven Diffusion (CVPR 2025 · similarity 0.84)
- Going Beyond Multi-Task Dense Prediction with Synergy Embedding Models (CVPR 2024 · similarity 0.84)
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.