Research map / Detection and segmentation / Open-set and few-shot detection
LawDIS: Language-Window-based Controllable Dichotomous Image Segmentation
Xinyu Yan, Meijun Sun, Ge-Peng Ji, Fahad Shahbaz Khan, Salman Khan, Deng-Ping Fan
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 joint macro/micro training with mode switcher (ablation) | LawDIS · DIS-TE4 | max F-measure: 0.912 -> 0.926 | Improved |
| fine-tune VAE decoder (ablation) | LawDIS · DIS-TE4 | max F-measure: 0.919 -> 0.926 | Improved |
| add language prompts in training and testing (ablation) | LawDIS · DIS-TE4 | max F-measure: 0.912 -> 0.926 | Improved |
| add language prompts at test time to prompt-trained model (ablation) | LawDIS · DIS-TE4 | max F-measure: 0.915 -> 0.926 | Improved |
| initialise window refinement from initial mask instead of Gaussian noise (ablation) | LawDIS Ours-R · DIS-TE4 | weighted F-measure: 0.843 -> 0.890 | Improved |
| add semi-auto window refinement (micro mode) to Ours-S (ablation) | LawDIS · DIS-TE4 | weighted F-measure: +2.0% | Improved |
| use one-step TCD scheduler instead of DDPM 500 steps (ablation) | LawDIS Ours-S · DIS-TE4 | FPS: 0.006 -> 3.09 | Improved |
| use one-step TCD instead of DDIM 50 steps (ablation) | LawDIS Ours-S · DIS-TE4 | weighted F-measure: 0.842 -> 0.856 | Improved |
| fine-tune VAE with TCD one-step (ablation) | LawDIS Ours-S · DIS-TE4 | weighted F-measure: 0.856 -> 0.890 | Improved |
| apply WR strategy as post-refinement to IS-Net (ablation) | IS-Net · DIS-TE4 | weighted F-measure: +6.4% | Improved |
| apply WR strategy as post-refinement to MVANet (ablation) | MVANet · DIS-TE4 | weighted F-measure: +3.2% | Improved |
| feed local patches to existing DIS models | BiRefNet · DIS-TE4 | weighted F-measure: -3.9% | Hurt |
Closest papers by meaning
- MaskFactory: Towards High-quality Synthetic Data Generation for Dichotomous Image Segmentation (NeurIPS 2024 · similarity 0.89)
- High-Precision Dichotomous Image Segmentation via Probing Diffusion Capacity (ICLR 2025 · similarity 0.88)
- $\mathbf{D^{3}S^{2}}$: Diffusion-Guided Dataset Distillation for Semantic Segmentation (NeurIPS 2026 · similarity 0.86)
- Explore In-Context Segmentation via Latent Diffusion Models (AAAI 2025 · similarity 0.86)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.86)
- ALTo: Adaptive-Length Tokenizer for Autoregressive Mask Generation (NeurIPS 2025 · 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.
