Research map / Detection and segmentation / CNN features
Searching for Efficient Multi-Scale Architectures for Dense Image Prediction
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 the searched DPC instead of DeepLabv3+ | Modified Xception dense prediction model · Cityscapes test | mIOU: 82.1 -> 82.7 | Helped |
| Use the searched DPC instead of the strongest listed prior model | Modified Xception dense prediction model · PASCAL-Person-Part validation | mIOU: 67.60 -> 71.34 | Helped |
| Use the searched DPC instead of MSCI | Modified Xception dense prediction model · PASCAL VOC 2012 test | mIOU: 88.0 -> 87.9 | Hurt |
| Replace ASPP with DPC | MobileNet-v2 model · Cityscapes validation | mIOU: 73.97 -> 75.38 | Helped |
| Replace ASPP with DPC | Modified Xception model · Cityscapes validation | mIOU: 80.25 -> 80.85 | Helped |
| Replace ASPP with DPC | Modified Xception model · Cityscapes validation | Params: 1.59M -> 0.81M | Helped |
| Replace ASPP with DPC | Modified Xception model · Cityscapes validation | MAdds: 18.12B -> 6.84B | Helped |
| Use MobileNet-v2 instead of the larger Xception backbone for the proxy (ablation) | Architecture ranking on Cityscapes · Cityscapes proxy task | Spearman rank correlation: ρ = 0.36 | Mixed |
| Cache backbone activations instead of training end-to-end (ablation) | MobileNet-v2 architecture-search proxy · Cityscapes proxy task | Spearman rank correlation: ρ = 0.47 | Helped |
| Train proxy candidates for 30K iterations rather than to convergence (ablation) | DPC architecture-search proxy · Dense prediction architecture ranking | runtime: 1+ week -> 90 minutes; 100+-fold | Helped |
Closest papers by meaning
- Fast Neural Architecture Search of Compact Semantic Segmentation Models via Auxiliary Cells (CVPR 2019 · similarity 0.88)
- DCNAS: Densely Connected Neural Architecture Search for Semantic Image Segmentation (CVPR 2021 · similarity 0.88)
- Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation (CVPR 2019 · similarity 0.87)
- Rethinking Semantic Segmentation From a Sequence-to-Sequence Perspective With Transformers (CVPR 2021 · similarity 0.86)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.85)
- Reinforcement Learning as a Parsimonious Alternative to Prediction Cascades: A Case Study on Image Segmentation (AAAI 2024 · similarity 0.85)
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.