Research map / Detection and segmentation / CNN features
InverseForm: A Loss Function for Structured Boundary-Aware 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 InverseForm to SA-Gate training | SA-Gate, ResNet-101 · NYU-Depth-v2 | mIoU: 52.4 -> 53.1 | Helped |
| Add InverseForm to HRNet-w18 training | HRNet-w18 · NYU-Depth-v2 | mIoU: 33.18 -> 34.79 | Helped |
| Add InverseForm to HRNet-w48 training | HRNet-w48 · NYU-Depth-v2 | mBA: 56.01 -> 59.34 | Helped |
| Add InverseForm to ASTMT boundary-loss training | SE-ResNet-101 + ASTMT · PASCAL-Context | Seg mIoU: 68.51 -> 70.14 | Helped |
| Add InverseForm to ASTMT boundary-loss training | SE-ResNet-101 + ASTMT · PASCAL-Context | Normals mean error: 14.37 -> 14.55 | Hurt |
| Add InverseForm to HRNet48 HMS training | HMS, HRNet48 · Cityscapes test | mIoU: 85.1 -> 85.6 | Helped |
| Add InverseForm to GSCNN training | GSCNN, WRN38 · Cityscapes val | mIoU: 81.0 -> 82.6 | Helped |
| Replace the AET convolutional loss network with the dense inverse-transformation network (ablation) | HRNet-w48 + InverseForm · NYU-Depth-v2 | mIoU: 47.03 -> 47.42 | Helped |
| Replace the AET convolutional loss network with the dense inverse-transformation network (ablation) | HRNet-w18 + InverseForm · NYU-Depth-v2 | mIoU: 33.97 -> 34.79 | Helped |
| Use Euclidean rather than geodesic distance (ablation) | HRNet-w48 + InverseForm · NYU-Depth-v2 | mIoU: 47.28 -> 47.42 | Mixed |
| Use Euclidean rather than geodesic distance (ablation) | HRNet-w18 + InverseForm · NYU-Depth-v2 | mIoU: 34.84 -> 34.79 | Mixed |
Closest papers by meaning
- Active Boundary Loss for Semantic Segmentation (AAAI 2022 · similarity 0.91)
- Beyond Pixel-wise Supervision: Local Structure Regularization for Semantic Segmentation (NeurIPS 2026 · similarity 0.88)
- Rethinking Semantic Segmentation From a Sequence-to-Sequence Perspective With Transformers (CVPR 2021 · similarity 0.87)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.86)
- Structural Pruning via Spatial-aware Information Redundancy for Semantic Segmentation (AAAI 2025 · similarity 0.86)
- Complex Wavelet Mutual Information Loss: A Multi-Scale Loss Function for Semantic Segmentation (ICML 2025 · 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.