Research map / Image generation / Autoregressive and latent generation
Symmetrical Flow Matching: Unified Image Generation, Segmentation, and Classification with Score-Based Generative Models
Francisco Caetano, Christiaan Viviers, Peter H.N. de With, Fons van der Sommen
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 |
|---|---|---|---|
| increase inference steps from 1 to 25 for image synthesis (ablation) | SymmFlow · CelebAMask-HQ | FID: 88.5 -> 11.9 | Improved |
| increase inference steps from 1 to 25 for image synthesis (ablation) | SymmFlow · COCO-Stuff | FID: 102.6 -> 7.0 | Improved |
| increase classification steps from 1 to 25 (ablation) | SymmFlow · CIFAR-10 classification | accuracy (%): 88.2 -> 90.6 | Improved |
| increase segmentation steps from 2 to 25 on CelebAMask-HQ (ablation) | SymmFlow · CelebAMask-HQ segmentation | mIoU: 70.3 -> 69.3 | No clear effect |
| increase segmentation steps from 1 to 20 on COCO-Stuff (ablation) | SymmFlow · COCO-Stuff segmentation | mIoU: 29.3 -> 40.1 | Improved |
| increase reverse-flow integration steps in toy classification (ablation) | SymmFlow MLP · Toy spiral dataset | accuracy (%): 100.0 -> 82.0 | Hurt |
| joint symmetric flow model instead of SemFlow | SymmFlow · COCO-Stuff semantic image synthesis | FID: 90.0 -> 7.0 | Improved |
| joint symmetric flow model instead of SemFlow | SymmFlow · COCO-Stuff segmentation | mIoU: 35.7 -> 39.6 | Improved |
| joint symmetric flow model instead of SemFlow | SymmFlow · CelebAMask-HQ segmentation | mIoU: 69.4 -> 69.3 | No clear effect |
| unified SymmFlow vs specialized segmenters | SymmFlow · COCO-Stuff segmentation | mIoU: 46.7 -> 39.6 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SCP-Diff — SCP-Diff: Spatial-Categorical Joint Prior for Diffusion Based Semantic Image Synthesis (ECCV 2024) · 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
- SemFlow: Binding Semantic Segmentation and Image Synthesis via Rectified Flow (NeurIPS 2024 · similarity 0.90)
- Blockwise Flow Matching: Improving Flow Matching Models For Efficient High-Quality Generation (NeurIPS 2025 · similarity 0.88)
- Rethinking Vector Field Learning for Generative Segmentation (NeurIPS 2026 · similarity 0.87)
- Improving Flow Matching by Aligning Flow Divergence (ICML 2025 · similarity 0.87)
- $\mathbf{D^{3}S^{2}}$: Diffusion-Guided Dataset Distillation for Semantic Segmentation (NeurIPS 2026 · similarity 0.86)
- Structured Coupling for Flow Matching (NeurIPS 2026 · 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.
