Research map / Detection and segmentation / Open-set and few-shot detection
FocalClick-XL: Toward Unified and High-Quality Interactive Segmentation
Xi Chen, Hengshuang Zhao
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 Target Crop to naive full-image model (ablation) | FocalClick-B0-S1 · DAVIS | NoC90: 15.60 -> 9.56 | Improved |
| add Focus Crop on top of Target Crop (ablation) | FocalClick-B0-S1 · DAVIS | NoC90: 9.56 -> 7.66 | Improved |
| add Progressive Merge (ablation) | FocalClick-B0-S1 · DAVIS-585 | NoC90: 8.52 -> 3.69 | Improved |
| add Progressive Merge (ablation) | FocalClick-B0-S2 · DAVIS | NoC90: 5.55 -> 5.49 | Improved |
| add Target Crop to naive model at larger input (ablation) | FocalClick-B0-S2 · DAVIS | NoC90: 9.74 -> 5.86 | Improved |
| add Object-Net to Context-Net (SAM) (ablation) | FocalClick-XL · DAVIS | NoC90: 5.12 -> 5.01 | Improved |
| add Detail-Net to Context-Net + Object-Net (ablation) | FocalClick-XL · DAVIS | NoC90: 5.01 -> 4.63 | Improved |
| add adapter after each Context-Net stage instead of none (ablation) | FocalClick-XL Context-Net · DAVIS | NoC90: 5.33 -> 4.63 | Improved |
| place adapter only at last layer instead of each stage (ablation) | FocalClick-XL Context-Net · GrabCut | NoC90: 1.48 -> 1.72 | Hurt |
| tune only prompting layer instead of full Object-Net when transferring click to scribble (ablation) | FocalClick-XL · DAVIS | NoC90 (scribble): 1.28 -> 1.35 | No clear effect |
| tune only Object-Net decoder instead of prompting layer (ablation) | FocalClick-XL · ADE-full | NoC90 (scribble): 4.51 -> 4.71 | Hurt |
| use box prompt instead of single click | FocalClick-XL-B · DAVIS | mIoU: 78.88 -> 84.89 | Improved |
Closest papers by meaning
- NTClick: Achieving Precise Interactive Segmentation With Noise-tolerant Clicks (CVPR 2025 · similarity 0.90)
- RClicks: Realistic Click Simulation for Benchmarking Interactive Segmentation (NeurIPS 2024 · similarity 0.90)
- TETRIS: Towards Exploring the Robustness of Interactive Segmentation (AAAI 2024 · similarity 0.89)
- Order-aware Interactive Segmentation (ICLR 2025 · similarity 0.89)
- AGILE3D: Attention Guided Interactive Multi-object 3D Segmentation (ICLR 2024 · similarity 0.88)
- FocSAM: Delving Deeply into Focused Objects in Segmenting Anything (CVPR 2024 · similarity 0.87)
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.
