Research map / Image generation / Fast diffusion sampling
Conditional Diffusion for Interactive Segmentation
Xi Chen, Zhiyan Zhao, Feiwu Yu, Yilei Zhang, Manni Duan
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 |
|---|---|---|---|
| add Diversified Training to the baseline (ablation) | DeeplabV3+ 5-channel baseline · DAVIS | NoC@90: 8.42 -> 8.14 | Improved |
| add Feature Diffusion Module (ablation) | CDNet · DAVIS | NoC@90: 8.42 -> 7.64 | Improved |
| add all three components (DT, FDM, PDM) to baseline (ablation) | DeeplabV3+ baseline · DAVIS | NoC@90: 8.42 -> 6.66 | Improved |
| add auxiliary destination loss to baseline without FDM (ablation) | baseline · DAVIS | NoC@90: 8.42 -> 8.61 | No clear effect |
| replace FDM with plain non-local layer (no constraints) (ablation) | baseline · DAVIS | NoC@90: 8.42 -> 8.81 | Hurt |
| add both source and destination constraints (FDM) vs source constraint only (ablation) | FDM · DAVIS | NoC@90: 8.49 -> 7.64 | Improved |
| supervise the destination constraint with auxiliary loss (ablation) | FDM · DAVIS | NoC@90: 8.04 -> 7.64 | Improved |
| use destination map for both source and destination (Des2Des) instead of click-based source (ablation) | FDM · DAVIS | NoC@90: 7.64 -> 8.10 | Hurt |
| increase PDM Gaussian enhancement amplitude from 0 to 1 (ablation) | PDM · DAVIS | NoC@90: 6.89 -> 6.66 | Improved |
| increase amplitude from 1 to 2 (ablation) | PDM · DAVIS | NoC@90: 6.66 -> 6.72 | Hurt |
| increase PDM diffusion iterations from 0 to 10 (ablation) | PDM · DAVIS | NoC@90: 7.54 -> 6.66 | Improved |
| increase PDM iterations from 10 to 20 (ablation) | PDM · DAVIS | NoC@90: 6.66 -> 6.66 | No clear effect |
Closest papers by meaning
- NTClick: Achieving Precise Interactive Segmentation With Noise-tolerant Clicks (CVPR 2025 · similarity 0.88)
- RClicks: Realistic Click Simulation for Benchmarking Interactive Segmentation (NeurIPS 2024 · similarity 0.87)
- TETRIS: Towards Exploring the Robustness of Interactive Segmentation (AAAI 2024 · similarity 0.87)
- Repurposing Stable Diffusion Attention for Training-Free Unsupervised Interactive Segmentation (CVPR 2025 · similarity 0.87)
- Explore In-Context Segmentation via Latent Diffusion Models (AAAI 2025 · similarity 0.86)
- Not All Diffusion Model Activations Have Been Evaluated as Discriminative Features (NeurIPS 2024 · similarity 0.85)
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.
