Research map / Detection and segmentation / CNN features
Gated-SCNN: Gated Shape CNNs for Semantic Segmentation
Towaki Takikawa, David Acuna, Varun Jampani, Sanja Fidler
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 GCL shape stream with dual task loss (ablation) | DeepLabV3+ ResNet-50 · Cityscapes val | mIoU: 71.3 -> 72.9 | Improved |
| add GCL shape stream with dual task loss (ablation) | DeepLabV3+ ResNet-101 · Cityscapes val | mIoU: 72.7 -> 74.3 | Improved |
| add GCL shape stream with dual task loss (ablation) | DeepLabV3+ Wide-ResNet · Cityscapes val | mIoU: 79.2 -> 79.8 | Improved |
| add GCL shape stream with dual task loss (ablation) | DeepLabV3+ ResNet-50 · Cityscapes val | boundary F-score th=5px: 68.5 -> 71.7 | Improved |
| add GCL shape stream with dual task loss (ablation) | DeepLabV3+ Wide-ResNet · Cityscapes val | boundary F-score th=5px: 73.0 -> 75.9 | Improved |
| add Canny image gradients before the fusion layer (ablation) | GSCNN ResNet-101 · Cityscapes val | mIoU: 74.3 -> 74.7 | Improved |
| add Canny image gradients before the fusion layer (ablation) | GSCNN ResNet-50 · Cityscapes val | mIoU: 72.9 -> 73.0 | No clear effect |
| add Canny image gradients before the fusion layer (ablation) | GSCNN ResNet-101 · Cityscapes val | boundary F-score th=5px: 73.3 -> 73.0 | Hurt |
| add dual task loss to GCL shape stream (ablation) | GSCNN ResNet-101 · Cityscapes val | boundary F-score th=3px: 65.0 -> 68.0 | Improved |
| add dual task loss to GCL shape stream (ablation) | GSCNN ResNet-101 · Cityscapes val | boundary F-score th=5px: 70.8 -> 73.0 | Improved |
| add GCL without dual task loss over baseline (ablation) | DeepLabV3+ ResNet-101 · Cityscapes val | boundary F-score th=3px: 64.1 -> 65.0 | Improved |
| replace DeepLabV3+ with GSCNN | DeepLabV3+ · Cityscapes val | mIoU: 78.8 -> 80.8 | Improved |
Closest papers by meaning
- Hierarchically Gated Deep Networks for Semantic Segmentation (CVPR 2016 · similarity 0.89)
- Efficient Piecewise Training of Deep Structured Models for Semantic Segmentation (CVPR 2016 · similarity 0.88)
- Gated Fully Fusion for Semantic Segmentation (AAAI 2020 · similarity 0.87)
- ShapeConv: Shape-Aware Convolutional Layer for Indoor RGB-D Semantic Segmentation (ICCV 2021 · similarity 0.87)
- DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs (TPAMI 2017 · similarity 0.87)
- SCTNet: Single-Branch CNN with Transformer Semantic Information for Real-Time Segmentation (AAAI 2024 · similarity 0.87)
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.
