Research map / Detection and segmentation / Open-set and few-shot detection
Universal Sketch Perceptual Grouping
Ke Li, Kaiyue Pang, Jifei Song, Yi-Zhe Song, Tao Xiang, Timothy M. Hospedales, Honggang Zhang
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 the global grouping loss (ablation) | Ours · SPG seen categories | VOI (lower is better): 0.63 -> 0.59 | Improved |
| add the global grouping loss (ablation) | Ours · SPG unseen categories | VOI (lower is better): 0.71 -> 0.64 | Improved |
| add the local grouping loss (ablation) | Ours · SPG seen categories | VOI (lower is better): 0.75 -> 0.59 | Improved |
| add the local grouping loss (ablation) | Ours · SPG unseen categories | VOI (lower is better): 0.95 -> 0.64 | Improved |
| add the reconstruction (generative) loss (ablation) | Ours · SPG seen categories | VOI (lower is better): 0.68 -> 0.59 | Improved |
| add the reconstruction (generative) loss (ablation) | Ours · SPG unseen categories | VOI (lower is better): 0.86 -> 0.64 | Improved |
| add the reconstruction (generative) loss (ablation) | Ours · SPG unseen categories | segmentation covering (SC): 0.69 -> 0.77 | Improved |
| add an l2 loss on the affinity matrix in place of separate local and global grouping losses (ablation) | Ours · SPG seen categories | VOI (lower is better): 0.59 -> 2.68 | Hurt |
| use the proposed grouper instead of Edge-PG | Ours · SPG unseen categories | VOI (lower is better): 1.13 -> 0.64 | Improved |
| use the proposed grouper instead of Edge-PG | Ours · Edge-PG datasets (unseen dataset) | VOI (lower is better): 1.69 -> 0.96 | Improved |
| use the proposed grouper instead of DeepLab | Ours · SPG seen categories | VOI (lower is better): 1.20 -> 0.59 | Improved |
| train FG-SBIR on sketches synthesised with the proposed grouper instead of LDSA | FG-SBIR pipeline · FG-SBIR Shoe-V2 | top-1 accuracy: 21.17% -> 26.88% | Improved |
Closest papers by meaning
- Generalising Fine-Grained Sketch-Based Image Retrieval (CVPR 2019 · similarity 0.87)
- Sketchformer: Transformer-Based Representation for Sketched Structure (CVPR 2020 · similarity 0.86)
- Sketch Me That Shoe (CVPR 2016 · similarity 0.86)
- Learning Geometry-Aware Representations by Sketching (CVPR 2023 · similarity 0.86)
- SketchFusion: Learning Universal Sketch Features through Fusing Foundation Models (CVPR 2025 · similarity 0.85)
- Perceptual Group Tokenizer: Building Perception with Iterative Grouping (ICLR 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.
