Research map / Detection and segmentation / Open-set and few-shot detection
Deep Hierarchical 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 hierarchical focal tree-min and tree-triplet losses to the CCE baseline (ablation) | HSSN with DeepLabV3+, ResNet-101 · Mapillary Vistas 2.0 val | mIoU^1: 37.43 -> 40.16 | Helped |
| Add focal tree-min loss (ablation) | HSSN with DeepLabV3+, ResNet-101 · Mapillary Vistas 2.0 val | mIoU^1: 37.43 -> 39.17 | Helped |
| Add tree-triplet loss (ablation) | HSSN with DeepLabV3+, ResNet-101 · PASCAL-Person-Part test | mIoU^1: 67.84 -> 72.27 | Helped |
| Use focal tree-min loss instead of tree-min loss (ablation) | HSSN with DeepLabV3+, ResNet-101 · PASCAL-Person-Part test | mIoU^1: 72.17 -> 72.89 | Helped |
| Set focal parameter γ to 5.0 instead of 2.0 (ablation) | HSSN with DeepLabV3+, ResNet-101 · Mapillary Vistas 2.0 val | mIoU^1: 40.16 -> 39.40 | Hurt |
| Use hierarchy-aware triplet samples instead of vanilla triplet samples (ablation) | HSSN with DeepLabV3+, ResNet-101 · Mapillary Vistas 2.0 val | mIoU^1: 39.41 -> 39.67 | Helped |
| Use hierarchy-induced margin instead of constant triplet margin (ablation) | HSSN with DeepLabV3+, ResNet-101 · PASCAL-Person-Part test | mIoU^1: 74.83 -> 75.44 | Helped |
| Use cosine instead of Euclidean distance for triplet learning (ablation) | HSSN with DeepLabV3+, ResNet-101 · Mapillary Vistas 2.0 val | mIoU^1: 39.33 -> 40.16 | Helped |
| Use cosine-annealed β instead of constant β=0.5 (ablation) | HSSN · Cityscapes val | mIoU^1: 82.68 -> 82.99 | Helped |
| Increase triplet count from 200 to 500 (ablation) | HSSN with DeepLabV3+, ResNet-101 · Mapillary Vistas 2.0 val | mIoU: 40.16 -> 40.19 | No clear effect |
| Apply HSSN to MaskFormer with Swin-Small instead of using MaskFormer alone | MaskFormer, Swin-Small · Mapillary Vistas 2.0 val | mIoU^1: 42.16 -> 43.97 | Helped |
| Apply HSSN to MaskFormer with Swin-Small instead of using MaskFormer alone | MaskFormer, Swin-Small · Cityscapes val | mIoU^1: 82.57 -> 83.74 | Helped |
Closest papers by meaning
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.91)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.90)
- Flattening the Parent Bias: Hierarchical Semantic Segmentation in the Poincare Ball (CVPR 2024 · similarity 0.90)
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (AAAI 2024 · similarity 0.87)
- Integrating Low-Level Visual Cues for Enhanced Unsupervised Semantic Segmentation (AAAI 2025 · similarity 0.87)
- Improving Visual Recognition with Hyperbolical Visual Hierarchy Mapping (CVPR 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.