Research map / Detection and segmentation / CNN features
Collaborative Global-Local Networks for Memory-Efficient Segmentation of Ultra-High Resolution Images
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 |
|---|---|---|---|
| Combine global and local branches through feature aggregation and feature-map regularization (ablation) | GLNet · DeepGlobe | mIoU (%): 57.3 and 66.4 -> 69.3; 69.3 -> 70.3 | Helped |
| Add shallow global-to-local feature sharing (ablation) | GLNet · DeepGlobe | mIoU (%): 70.3 -> 70.5 | Helped |
| Replace shallow global-to-local sharing with deep sharing (ablation) | GLNet · DeepGlobe | mIoU (%): 70.5 -> 70.9 | Helped |
| Add local-to-global sharing to deep global-to-local sharing (ablation) | GLNet · DeepGlobe | mIoU (%): 70.9 -> 71.6 | Helped |
| Use only one branch instead of global and local information together (ablation) | GLNet · DeepGlobe | mIoU (%): 57.3; 66.4 -> 71.6 | Hurt |
| Add coarse-to-fine bounding-box refinement (ablation) | GLNet · ISIC | mIoU (%): 72.7 -> 73.9 | Helped |
| Add local-to-global sharing to coarse-to-fine global-to-local sharing (ablation) | Coarse-to-fine GLNet · ISIC | mIoU (%): 73.9 -> 75.2 | Helped |
| Use only the global branch instead of the coarse-to-fine bidirectional model (ablation) | Coarse-to-fine GLNet · ISIC | mIoU (%): 70.1 -> 75.2 | Hurt |
| Use only the local branch instead of the coarse-to-fine bidirectional model (ablation) | Coarse-to-fine GLNet · ISIC | mIoU (%): 48.5 -> 75.2 | Hurt |
| Add global-to-local feature sharing (ablation) | GLNet · Inria Aerial | mIoU (%): 63.1 -> 66.0 | Helped |
| Add local-to-global sharing to global-to-local sharing (ablation) | GLNet · Inria Aerial | mIoU (%): 66.0 -> 71.2 | Helped |
| Use only the global branch instead of bidirectional global-local collaboration (ablation) | GLNet · Inria Aerial | mIoU (%): 42.5 -> 71.2 | Hurt |
Closest papers by meaning
- Patch Proposal Network for Fast Semantic Segmentation of High-Resolution Images (AAAI 2020 · similarity 0.89)
- ISDNet: Integrating Shallow and Deep Networks for Efficient Ultra-High Resolution Segmentation (CVPR 2022 · similarity 0.87)
- xT: Nested Tokenization for Larger Context in Large Images (ICML 2024 · similarity 0.86)
- Toward Real Ultra Image Segmentation: Leveraging Surrounding Context to Cultivate General Segmentation Model (NeurIPS 2024 · similarity 0.86)
- Ultra-High Resolution Segmentation via Boundary-Enhanced Patch-Merging Transformer (AAAI 2025 · similarity 0.85)
- ALGM: Adaptive Local-then-Global Token Merging for Efficient Semantic Segmentation with Plain Vision Transformers (CVPR 2024 · similarity 0.84)
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.