Research map / Graph learning / Graph self-supervised learning
Semi-Supervised Learning for Multi-Task Scene Understanding by Neural Graph Consensus
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 |
|---|---|---|---|
| Train EdgeNets on consensus pseudo-labels through the multi-path Neural Graph Consensus graph | EdgeNet scene-interpretation predictors · Depth (evaluation set) | L1 (meters): 4.9844 -> 3.2994 | Helped |
| Train EdgeNets on consensus pseudo-labels through the multi-path Neural Graph Consensus graph | EdgeNet scene-interpretation predictors · Semantic Segmentation (evaluation set) | mIOU: 0.4840 -> 0.5258 | Helped |
| Train EdgeNets on consensus pseudo-labels through the multi-path Neural Graph Consensus graph | EdgeNet scene-interpretation predictors · Position (evaluation set) | L2 (meters): 25.7597 -> 12.0764 | Helped |
| Train EdgeNets on consensus pseudo-labels through the multi-path Neural Graph Consensus graph | EdgeNet scene-interpretation predictors · Orientation (evaluation set) | L1 (degrees): 3.8439 -> 2.2088 | Helped |
| Add a second iteration of consensus learning on new unlabeled data (ablation) | NGC · Semantic Segmentation (evaluation set) | mIOU: 0.4978 -> 0.5258 | Helped |
| Add a second iteration of consensus learning on new unlabeled data (ablation) | NGC · Depth (evaluation set) | L1 (meters): 3.4867 -> 3.2994 | Helped |
| Use a distilled single EdgeNet instead of the multi-path NGC ensemble (ablation) | NGC · Semantic Segmentation (evaluation set) | mIOU: 0.5258 -> 0.5159 | Hurt |
| Use a distilled single EdgeNet instead of the multi-path NGC ensemble (ablation) | NGC · Depth (evaluation set) | L1 (meters): 3.2994 -> 3.9508 | Hurt |
| Use NGC consensus learning rather than CCT semi-supervised learning | Semantic Segmentation · Evaluation set | mIOU: 0.353 -> 0.526 | Helped |
| Use NGC consensus learning rather than CCT semi-supervised learning | Semantic Segmentation · Evaluation set | Accuracy: 0.8463 -> 0.8503 | Helped |
| Use multi-representation graph consensus rather than a vanilla EdgeNets ensemble | Semantic Segmentation · Evaluation set | mIOU: 0.5248 -> 0.5258 | Helped |
| Use multi-representation graph consensus rather than a vanilla EdgeNets ensemble | Depth · Evaluation set | L1 (meters): 3.3509 -> 3.2994 | Helped |
Closest papers by meaning
- Taskology: Utilizing Task Relations at Scale (CVPR 2021 · similarity 0.87)
- Classification by Attention: Scene Graph Classification with Prior Knowledge (AAAI 2021 · similarity 0.87)
- Stacked Hybrid-Attention and Group Collaborative Learning for Unbiased Scene Graph Generation (CVPR 2022 · similarity 0.86)
- Big Self-Supervised Models are Strong Semi-Supervised Learners (NeurIPS 2020 · similarity 0.86)
- Relaxed Class-consensus Consistency for Semi-supervised Semantic Segmentation (AAAI 2025 · similarity 0.86)
- SURER: Structure-Adaptive Unified Graph Neural Network for Multi-View Clustering (AAAI 2024 · similarity 0.84)
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.