Research map / Detection and segmentation / CNN features
Graph-Guided Architecture Search 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 |
|---|---|---|---|
| Add coarse training data | GAS · Cityscapes test | mIoU (%): 71.8 -> 73.5 | Helped |
| Add communication between independent cells using GGM rather than a fully connected layer (ablation) | GAS · Cityscapes validation | mIoU (%): about 3% | Helped |
| Use independent cells instead of shared cells (ablation) | GAS · Cityscapes validation | mIoU (%) | Hurt |
| Use GCN-guided communication instead of random architecture sampling without a latency constraint (ablation) | GAS · Cityscapes validation | mIoU (%): 69.6 -> 72.3 | Helped |
| Use GCN-guided communication instead of random sampling under the speed requirement (ablation) | GAS · Cityscapes validation | mIoU (%): 65.8 -> 72.3 | Helped |
| Set GCN weight dimension to 64 rather than 16 (ablation) | GAS · Cityscapes validation | mIoU (%): 71.6 -> 72.4 | Helped |
| Set GCN weight dimension to 64 rather than 32 (ablation) | GAS · Cityscapes validation | mIoU (%): 71.8 -> 72.4 | Helped |
| Set GCN weight dimension to 128 rather than 64 (ablation) | GAS · Cityscapes validation | mIoU (%): 72.4 -> 72.1 | Hurt |
| Set GCN weight dimension to 256 rather than 64 (ablation) | GAS · Cityscapes validation | mIoU (%): 72.4 -> 71.5 | Hurt |
| Construct the reasoning graph from operation identities rather than edge similarities (ablation) | GAS · Cityscapes validation | mIoU (%): 72.4 -> 70.9 | Hurt |
| Increase the latency-loss weight from 0.0005 to 0.005 (ablation) | GAS · Cityscapes validation | mIoU and latency | Mixed |
| Increase the latency-loss weight from 0.005 to 0.05 (ablation) | GAS · Cityscapes validation | mIoU and latency | Mixed |
Closest papers by meaning
- Customizable Architecture Search for Semantic Segmentation (CVPR 2019 · similarity 0.91)
- Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation (CVPR 2019 · similarity 0.91)
- Fast Neural Architecture Search of Compact Semantic Segmentation Models via Auxiliary Cells (CVPR 2019 · similarity 0.90)
- FasterSeg: Searching for Faster Real-time Semantic Segmentation (ICLR 2020 · similarity 0.89)
- SCTNet: Single-Branch CNN with Transformer Semantic Information for Real-Time Segmentation (AAAI 2024 · similarity 0.87)
- SasWOT: Real-Time Semantic Segmentation Architecture Search WithOut Training (AAAI 2024 · similarity 0.86)
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.