Research map / Detection and segmentation / CNN features
Pyramid Scene Parsing Network
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 pyramid pooling with multiple regional scales (ablation) | ResNet50+DA+AL · ADE20K validation | Mean IoU(%): 37.23 -> 41.68 | Helped |
| Use four-level pooling instead of global 1×1 pooling (ablation) | ResNet50 with average pooling · ADE20K validation | Mean IoU(%): 40.07 -> 41.07 | Helped |
| Use average rather than max pooling (ablation) | ResNet50 pyramid pooling · ADE20K validation | Mean IoU(%): 40.18 -> 41.07 | Helped |
| Add dimension reduction after pooling (ablation) | ResNet50 with four-level average pooling · ADE20K validation | Mean IoU(%): 41.07 -> 41.68 | Helped |
| Add auxiliary loss with α = 0.4 (ablation) | ResNet50 FCN · ADE20K validation | Mean IoU(%): 35.82 -> 37.23 | Helped |
| Set auxiliary-loss weight to 0.3 instead of 0.4 (ablation) | ResNet50 FCN · ADE20K validation | Mean IoU(%): 37.23 -> 37.01 | Hurt |
| Set auxiliary-loss weight to 0.6 instead of 0.4 (ablation) | ResNet50 FCN · ADE20K validation | Mean IoU(%): 37.23 -> 37.09 | Hurt |
| Set auxiliary-loss weight to 0.9 instead of 0.4 (ablation) | ResNet50 FCN · ADE20K validation | Mean IoU(%): 37.23 -> 36.99 | Hurt |
| Increase the pretrained ResNet depth from 50 to 269 (ablation) | PSPNet · ADE20K validation | Mean IoU(%): 41.68 -> 43.81 | Helped |
| Add multi-scale testing (ablation) | PSPNet with ResNet269 · ADE20K validation | Mean IoU(%): 43.81 -> 44.94 | Helped |
| Add data augmentation (ablation) | ResNet50 baseline · ADE20K validation | Mean IoU(%): 34.28 -> 35.82 | Helped |
| Add auxiliary loss (ablation) | ResNet50+DA · ADE20K validation | Mean IoU(%): 35.82 -> 37.23 | Helped |
Closest papers by meaning
- Strip Pooling: Rethinking Spatial Pooling for Scene Parsing (CVPR 2020 · similarity 0.89)
- Adaptive Pyramid Context Network for Semantic Segmentation (CVPR 2019 · similarity 0.87)
- Scene Parsing Through ADE20K Dataset (CVPR 2017 · similarity 0.87)
- VSPW: A Large-scale Dataset for Video Scene Parsing in the Wild (CVPR 2021 · similarity 0.86)
- Deep Structured Scene Parsing by Learning With Image Descriptions (CVPR 2016 · similarity 0.85)
- 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.