Research map / Detection and segmentation / Open-set and few-shot detection
Scene Essence
Jiayan Qiu, Yiding Yang, Xinchao Wang, Dacheng Tao
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 hierarchical GNN clustering instead of GAT for object grouping | Scene Essence · Places365 | accuracy (%): 50.64 -> 55.21 | Improved |
| Use hierarchical GNN clustering instead of GAT for object grouping | Scene Essence · SUN397 | accuracy (%): 62.91 -> 68.31 | Improved |
| Use hierarchical GNN clustering instead of GAT for object grouping | Scene Essence · MIT67 | accuracy (%): 76.83 -> 83.92 | Improved |
| Add object bounding-box coordinates to object features (ablation) | Scene Essence · Places365 | accuracy (%): 43.84 -> 55.21 | Improved |
| Add object bounding-box coordinates to object features (ablation) | Scene Essence · SUN397 | accuracy (%): 54.58 -> 68.31 | Improved |
| Add object bounding-box coordinates to object features (ablation) | Scene Essence · MIT67 | accuracy (%): 60.37 -> 83.92 | Improved |
| Increase λ from 0.5 to 0.7 to penalize kept objects more (ablation) | Scene Essence · Places365 | accuracy (%): 55.21 -> 9.57 | Hurt |
| Increase λ from 0.5 to 0.9 to penalize kept objects more (ablation) | Scene Essence · Places365 | accuracy (%): 55.21 -> 6.09 | Hurt |
| Increase λ from 0.5 to 0.7 (ablation) | Scene Essence · Places365 | erasing ratio: 0.58 -> 0.64 | Improved |
| Increase λ from 0.5 to 0.9 (ablation) | Scene Essence · Places365 | erasing ratio: 0.58 -> 0.71 | Improved |
| Erase one additional kept object from the learned Scene Essence (ablation) | Scene Essence · Places365 | accuracy (%): 53.95 -> 17.11 | Hurt |
Closest papers by meaning
- A Comprehensive Survey of Scene Graphs: Generation and Application (TPAMI 2021 · similarity 0.87)
- Learning Physical Graph Representations from Visual Scenes (NeurIPS 2020 · similarity 0.86)
- Object-level Scene Context Prediction (TPAMI 2021 · similarity 0.86)
- Classification by Attention: Scene Graph Classification with Prior Knowledge (AAAI 2021 · similarity 0.86)
- Learning To Generate Scene Graph From Natural Language Supervision (ICCV 2021 · similarity 0.86)
- SCENIR: Visual Semantic Clarity through Unsupervised Scene Graph Retrieval (ICML 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.
