Research map / Detection and segmentation / Open-set and few-shot detection
FlowDIS: Language-Guided Dichotomous Image Segmentation with Flow Matching
Andranik Sargsyan, Shant Navasardyan
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 |
|---|---|---|---|
| use deterministic flow matching from image latent instead of Gaussian noise (ablation) | FlowDIS · DIS-VD | weighted F-measure: 0.883 -> 0.938 | Improved |
| use deterministic flow matching from image latent instead of Gaussian noise (ablation) | FlowDIS · DIS-VD | MAE: 0.025 -> 0.014 | Improved |
| add language guidance (ablation) | FlowDIS · DIS-VD | weighted F-measure: 0.901 -> 0.937 | Improved |
| add language guidance (ablation) | FlowDIS · DIS-VD | MAE: 0.027 -> 0.015 | Improved |
| add PAIP training (ablation) | FlowDIS · DIS-VD-Complex | max F-measure: 0.783 -> 0.960 | Improved |
| add PAIP training (ablation) | FlowDIS · DIS-VD-Complex | MAE: 0.063 -> 0.014 | Improved |
| add PAIP training (ablation) | FlowDIS · DIS-VD (simple scenes) | max F-measure: 0.956 -> 0.958 | No clear effect |
| add PAIP training (ablation) | FlowDIS · COCO-Object (zero-shot) | weighted F-measure: 0.327 -> 0.511 | Improved |
| add PAIP training (ablation) | FlowDIS · COCO-Object (zero-shot) | MAE: 0.191 -> 0.075 | Improved |
| concatenate image latent z^I to the input (ablation) | FlowDIS · DIS-TE (1-4) | weighted F-measure: 0.933 -> 0.938 | Improved |
| concatenate image latent z^I to the input (ablation) | FlowDIS · DIS-TE (1-4) | MAE: 0.017 -> 0.016 | Improved |
| increase inference resolution from 1024 to 2048 px | FlowDIS · DIS-TE4 | weighted F-measure: 0.919 -> 0.932 | Improved |
Closest papers by meaning
- High-Precision Dichotomous Image Segmentation via Probing Diffusion Capacity (ICLR 2025 · similarity 0.87)
- Multi-view Aggregation Network for Dichotomous Image Segmentation (CVPR 2024 · similarity 0.85)
- $\mathbf{D^{3}S^{2}}$: Diffusion-Guided Dataset Distillation for Semantic Segmentation (NeurIPS 2026 · similarity 0.83)
- MaskFactory: Towards High-quality Synthetic Data Generation for Dichotomous Image Segmentation (NeurIPS 2024 · similarity 0.83)
- DiffCut: Catalyzing Zero-Shot Semantic Segmentation with Diffusion Features and Recursive Normalized Cut (NeurIPS 2024 · similarity 0.83)
- Rethinking Vector Field Learning for Generative Segmentation (NeurIPS 2026 · 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.
