Research map / Detection and segmentation / CNN features
Dynamic Sampling Network for Semantic Segmentation
Read the paperCheck 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 |
|---|---|---|---|
| use low-level feature X(3) instead of X(4) with ASPP high-level (ablation) | CGDS module · PASCAL VOC 2012 val | mIoU (%): 76.7 -> 79.0 | Helped |
| use low-level X(3) instead of X(5) with ASPP (ablation) | CGDS module · PASCAL VOC 2012 val | mIoU (%): 78.1 -> 79.0 | Helped |
| use ASPP multi-scale context instead of X(5) as high-level feature (ablation) | CGDS module · PASCAL VOC 2012 val | mIoU (%): 77.4 -> 79.0 | Helped |
| replace conv sampling with non-local sub-network (ablation) | CGDS module · PASCAL VOC 2012 val | mIoU (%): 79.0 -> 76.6 | Hurt |
| compare with Deformable Convolution v2 on ASPP (ablation) | DSNet · PASCAL VOC 2012 val | mIoU (%): 75.4 -> 79.0 | Helped |
| compare with Dynamic Filter on ASPP (ablation) | DSNet · PASCAL VOC 2012 val | mIoU (%): 74.6 -> 79.0 | Helped |
| add deep supervision with CGDS module (DSNet vs baseline) (ablation) | Baseline · PASCAL VOC 2012 val | Mean IoU (%): 80.06 -> 80.98 | Helped |
| add horizontal flip at evaluation (ablation) | DSNet · PASCAL VOC 2012 val | Mean IoU (%): 80.98 -> 81.63 | Helped |
| add multi-scale inputs at inference (ablation) | DSNet · PASCAL VOC 2012 val | Mean IoU (%): 81.63 -> 82.53 | Helped |
| fine-tune on original VOC training set (ablation) | DSNet · PASCAL VOC 2012 val | Mean IoU (%): 82.53 -> 83.74 | Helped |
| CGDS module versus DeepLab v3+ | DSNet · PASCAL VOC 2012 val | mIoU (%): +1.14% | Helped |
| DSNet vs best prior on PASCAL Context (APCNet) | DSNet · PASCAL Context | mIoU (%): 54.7 -> 54.2 | No clear effect |
Closest papers by meaning
- Context Encoding for Semantic Segmentation (CVPR 2018 · similarity 0.87)
- Context-Reinforced Semantic Segmentation (CVPR 2019 · similarity 0.87)
- Efficient Piecewise Training of Deep Structured Models for Semantic Segmentation (CVPR 2016 · similarity 0.87)
- Hierarchically Gated Deep Networks for Semantic Segmentation (CVPR 2016 · similarity 0.86)
- Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling (CVPR 2024 · similarity 0.86)
- Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation (NeurIPS 2025 · 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.