Research map / Label-efficient and robust learning / Semi-supervised learning
Hunting Sparsity: Density-Guided Contrastive Learning for Semi-Supervised Semantic Segmentation
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 |
|---|---|---|---|
| Apply DGCL rather than the strongest reported baseline | Semi-supervised segmentation with DeepLabv3+ and ResNet-101 · PASCAL VOC 2012 val, classic 1/8 (183) | mIoU (%): +2.43% | Helped |
| Apply DGCL rather than the strongest reported baseline | Semi-supervised segmentation with DeepLabv3+ and ResNet-101 · PASCAL VOC 2012 val, classic 1/4 (366) | mIoU (%): +1.57% | Helped |
| Apply DGCL rather than the strongest reported baseline | Semi-supervised segmentation with DeepLabv3+ and ResNet-101 · PASCAL VOC 2012 val, blended 1/16 (662) | mIoU (%): +1.11% | Helped |
| Apply DGCL rather than the strongest reported baseline | Semi-supervised segmentation with DeepLabv3+ and ResNet-101 · Cityscapes val, 1/2 (1488) | mIoU (%): +1.60% | Helped |
| Add density-guided sampling instead of random contrastive sampling and average-pooled centers (ablation) | Self-training segmentation · PASCAL VOC 2012, classic 1/8 (183) | mIoU (%): +4.01% | Helped |
| Add density-guided sampling instead of random contrastive sampling and average-pooled centers (ablation) | Self-training segmentation · PASCAL VOC 2012, blended 1/8 (1323) | mIoU (%): +1.24% | Helped |
| Add class memory for density estimation and memory features as positive keys (ablation) | Density-guided contrastive learning · PASCAL VOC 2012, classic 1/8 (183) | mIoU (%): +1.31% | Helped |
| Add class memory for density estimation and memory features as positive keys (ablation) | Density-guided contrastive learning · PASCAL VOC 2012, blended 1/8 (1323) | mIoU (%): +0.40% | Helped |
| Select anchors and keys by prediction confidence rather than feature density (ablation) | Contrastive learning for semi-supervised segmentation · PASCAL VOC 2012, classic 1/8 (183) | mIoU (%): 71.42 | Hurt |
| Use a multi-scale graph with 8, 16, and 32 neighbors rather than a single-scale graph (ablation) | DGCL density estimation · PASCAL VOC 2012, classic 1/8 (183) | mIoU (%) | Helped |
Closest papers by meaning
- Towards the Uncharted: Density-Descending Feature Perturbation for Semi-supervised Semantic Segmentation (CVPR 2024 · similarity 0.91)
- Neighbor Does Matter: Density-Aware Contrastive Learning for Medical Semi-supervised Segmentation (AAAI 2025 · similarity 0.91)
- Density-Guided Semi-Supervised 3D Semantic Segmentation with Dual-Space Hardness Sampling (CVPR 2024 · similarity 0.88)
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.87)
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (AAAI 2024 · similarity 0.87)
- RankMatch: Exploring the Better Consistency Regularization for Semi-supervised Semantic Segmentation (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.