Research map / Detection and segmentation / CNN features
DFANet: Deep Feature Aggregation for Real-Time Semantic Segmentation
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 |
|---|---|---|---|
| Replace ResNet-50 with Xception A (ablation) | Cityscapes val segmentation backbone · Cityscapes val | mIoU: 68.3% -> 59.2% | Hurt |
| Add ASPP to Xception A (ablation) | Xception A · Cityscapes val | mIoU: 59.2% -> 67.1% | Helped |
| Add FC attention to Xception A (ablation) | Xception A · Cityscapes val | mIoU: 59.2 -> 65.4 | Helped |
| Add one additional backbone for sub-network aggregation (ablation) | Backbone A · Cityscapes val | mIoU: 65.4% -> 66.3% | Helped |
| Add a third Backbone A encoder (ablation) | Backbone A x2 · Cityscapes val | mIoU: 66.3% -> 65.1% | Hurt |
| Add a fourth Backbone A encoder (ablation) | Backbone A x3 · Cityscapes val | mIoU: 65.1% -> 50.8% | Hurt |
| Add high-level feature fusion (ablation) | Backbone A x3 · Cityscapes val | mIoU: 65.1% -> 69.6% | Helped |
| Add low-level feature fusion after high-level fusion (ablation) | Backbone A x3 + HL · Cityscapes val | mIoU: 69.6% -> 71.9% | Helped |
| Add high-level feature fusion (ablation) | Backbone B x3 · Cityscapes val | mIoU: 58.2% -> 67.6% | Helped |
| Add low-level feature fusion after high-level fusion (ablation) | Backbone B x3 + HL · Cityscapes val | mIoU: 67.6% -> 68.4% | Helped |
| Reduce Backbone A input height to 512 × 1024 (ablation) | DFANet A · Cityscapes test | mIoU: 71.3 -> 70.3 | Hurt |
| Use the lighter Backbone B instead of Backbone A (ablation) | DFANet · Cityscapes test | mIoU: 71.3 -> 67.1 | Hurt |
Closest papers by meaning
- Rethinking BiSeNet for Real-Time Semantic Segmentation (CVPR 2021 · similarity 0.88)
- Dynamic Video Segmentation Network (CVPR 2018 · similarity 0.88)
- SCTNet: Single-Branch CNN with Transformer Semantic Information for Real-Time Segmentation (AAAI 2024 · similarity 0.87)
- FasterSeg: Searching for Faster Real-time Semantic Segmentation (ICLR 2020 · similarity 0.86)
- SasWOT: Real-Time Semantic Segmentation Architecture Search WithOut Training (AAAI 2024 · similarity 0.85)
- Multi-view Aggregation Network for Dichotomous Image Segmentation (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.