Research map / Architectures and efficiency / CNN design
Strip Pooling: Rethinking Spatial Pooling for Scene Parsing
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 two MPMs to Base FCN (ablation) | Base FCN with ResNet-50 · ADE20K | mIoU: 37.63 -> 41.92 | Helped |
| Add a second MPM instead of using one (ablation) | Base FCN with ResNet-50 · ADE20K | mIoU: 40.50 -> 41.92 | Helped |
| Add more than two MPMs (ablation) | Base FCN with ResNet-50 · ADE20K | mIoU | No clear effect |
| Add long-range strip pooling to short-range aggregation (ablation) | Two-MPM Base FCN · ADE20K | mIoU: 40.50 -> 41.92 | Helped |
| Add short-range aggregation to long-range strip pooling (ablation) | Two-MPM Base FCN · ADE20K | mIoU: 41.14 -> 41.92 | Helped |
| Add SPM at the last building block of each stage (ablation) | Base FCN + 2 MPM · ADE20K | mIoU: 41.92 -> 42.61 | Helped |
| Add SPM to all blocks in the last stage rather than only the last block of each stage (ablation) | Base FCN + 2 MPM · ADE20K | mIoU: 42.61 -> 42.30 | Hurt |
| Add SPM at both the last block of each stage and all blocks in the last stage (ablation) | Base FCN + 2 MPM · ADE20K | mIoU: 41.92 -> 44.03 | Helped |
| Add SPMs to all backbone building blocks (ablation) | Base FCN + 2 MPM · ADE20K | mIoU | No clear effect |
| Replace strip pooling in SPM with global average pooling (ablation) | Base FCN + SPM · ADE20K | mIoU: 41.92 -> 41.34 | Hurt |
| Use ResNet-101 instead of ResNet-50 (ablation) | SPNet · ADE20K | mIoU: 45.03% -> 45.60% | Helped |
| Add multi-scale testing with flipping (ablation) | SPNet-50 · ADE20K | mIoU: 44.03 -> 45.03 | Helped |
Closest papers by meaning
- Pyramid Scene Parsing Network (CVPR 2017 · similarity 0.89)
- AttaNet: Attention-Augmented Network for Fast and Accurate Scene Parsing (AAAI 2021 · similarity 0.87)
- SceneSplat++: A Large Dataset and Comprehensive Benchmark for Language Gaussian Splatting (NeurIPS 2025 · similarity 0.84)
- Tracing Hyperparameter Dependencies for Model Parsing via Learnable Graph Pooling Network (NeurIPS 2024 · similarity 0.84)
- ResCLIP: Residual Attention for Training-free Dense Vision-language Inference (CVPR 2025 · similarity 0.84)
- The Scene Language: Representing Scenes with Programs, Words, and Embeddings (CVPR 2025 · similarity 0.83)
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.