Research map / Vision-language and multimodal / Human and embodied modeling
S2-Boost: Synergistic Semantic Boosting for Coarse-to-Fine Ensemble Learning
Guanxiong He, Zheng Wang, Jie Wang, Liaoyuan Tang, Rong Wang, Feiping Nie
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 boosting over semantic partitions (ablation) | S2-Boosting · ImageNet-1K | Top-1 accuracy: 76.73 -> 88.35 | Improved |
| use learned semantic masks instead of random grid masks (ablation) | S2-Boosting · ImageNet-1K | Top-1 accuracy: 85.07 -> 88.35 | Improved |
| add diversity and center losses to SPSLM (ablation) | S2-Boosting · ImageNet-1K | Top-1 accuracy: 83.35 -> 88.35 | Improved |
| add reconstruction loss to SBM (ablation) | S2-Boosting · ImageNet-1K | Top-1 accuracy: 81.37 -> 88.35 | Improved |
| order masks by mean activation (coarse-to-fine) instead of random order (ablation) | SBM · ImageNet-1K | Top-1 accuracy | Improved |
| use S2-Boosting instead of plain ViT-L | ViT-L · ImageNet-1K | Top-1 accuracy: 85.15 -> 88.35 | Improved |
| use S2-Boosting instead of Re-Boosting | ViT-L · ImageNet-1K | Top-1 accuracy: 84.71 -> 88.35 | Improved |
| use S2-Boosting instead of Re-Boosting | ViT-L · CIFAR-10 | Top-1 accuracy: 99.24 -> 99.51 | Improved |
| use S2-Boosting instead of MAE | ViT-L · CIFAR-100 | Top-1 accuracy: 93.31 -> 93.47 | Improved |
| apply S2-Boosting under 50% random patch occlusion | ViT-L · ImageNet-1K (50% occlusion) | Top-1 accuracy: 60.37 -> 75.61 | Improved |
| apply S2-Boosting under 20% random patch occlusion | ViT-L · ImageNet-1K (20% occlusion) | Top-1 accuracy: 77.88 -> 82.19 | Improved |
Closest papers by meaning
- L-WISE: Boosting Human Visual Category Learning Through Model-Based Image Selection and Enhancement (ICLR 2025 · similarity 0.85)
- Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks (CVPR 2024 · similarity 0.85)
- Boosting Segment Anything Model Towards Open-Vocabulary Learning (AAAI 2025 · similarity 0.85)
- Understanding Overadaptation in Supervised Fine-Tuning: The Role of Ensemble Methods (ICML 2025 · similarity 0.85)
- Modeling Collaborator: Enabling Subjective Vision Classification With Minimal Human Effort via LLM Tool-Use (CVPR 2024 · similarity 0.85)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 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.
