Research map / Image generation / Fast diffusion sampling
Diffusion Adversarial Representation Learning for Self-supervised Vessel Segmentation
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 diffusion module with L_diff (vs. without) (ablation) | DARL generation/segmentation · XCAD | IoU: 0.471 vs 0.449 (about 2 points) | Helped |
| add switchable SPADE layers (vs. plain generation without S-SPADE) (ablation) | generation module · XCAD | IoU: 0.439 without S-SPADE vs 0.449 with diffusion-only ablation; roughly 1 point+ | Helped |
| add cyclic reconstruction loss L_cyc (ablation) | DARL · XCAD | IoU: 0.471 vs 0.322 | Helped |
| use L1 instead of cross-entropy for cyclic loss (ablation) | L_cyc · XCAD | IoU: 0.471 vs 0.346 | Helped |
| replace diffusion latent features with autoencoder latent (ablation) | DARL latent estimator · XCAD | IoU: 0.471 vs 0.399 | Helped |
| remove discriminator D_s (ablation) | DARL · XCAD | IoU: 0.456 vs 0.471 (precision rises to 0.728 from 0.701) | Hurt |
| remove discriminator D_a (ablation) | DARL · XCAD | IoU: 0.348 vs 0.471 | Hurt |
| DARL vs. best baselines (self-supervised SSVS) on XCAD | DARL · XCAD | IoU: 0.471 vs 0.410 (SSVS) | Helped |
| DARL vs. baselines on external 134 XCA | DARL · 134 XCA | IoU: 0.426 vs 0.318 (SSVS) | Helped |
| DARL under Gaussian noise sigma=50 vs baselines | DARL · XCAD with sigma=50 noise | IoU: 0.269 vs 0.102 (DA) | Helped |
| use an unsupervised input normalization (HE/CLAHE) for baselines (ablation) | unsupervised baselines SGC, Redrawing, DS · XCAD | Dice: SGC 0.061 (HE) vs 0.111; CLAHE 0.119 vs 0.111; DS drops 0.526 to 0.492/0.495 | Mixed |
Closest papers by meaning
- vesselFM: A Foundation Model for Universal 3D Blood Vessel Segmentation (CVPR 2025 · similarity 0.88)
- DeNVeR: Deformable Neural Vessel Representations for Unsupervised Video Vessel Segmentation (CVPR 2025 · similarity 0.87)
- Learning Dynamic Collaborative Network for Semi-supervised 3D Vessel Segmentation (CVPR 2025 · similarity 0.85)
- VasTSD: Learning 3D Vascular Tree-state Space Diffusion Model for Angiography Synthesis (CVPR 2025 · similarity 0.84)
- Factorized Diffusion Architectures for Unsupervised Image Generation and Segmentation (NeurIPS 2024 · similarity 0.83)
- Noise-Consistent Siamese-Diffusion for Medical Image Synthesis and Segmentation (CVPR 2025 · similarity 0.83)
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.