Research map / Graph learning / Knowledge and temporal graphs
Graphonomy: Universal Image Parsing via Graph Reasoning and Transfer
Liang Lin, Yiming Gao, Ke Gong, Meng Wang, Xiaodan Liang
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 Intra-Graph Reasoning with adjacency matrix to basic network (ablation) | DeepLab v3+ · PASCAL-Person-Part | mean IoU (%): 67.84 -> 68.34 | Improved |
| add adjacency matrix A^e prior to Intra-Graph Reasoning (ablation) | Intra-Graph Reasoning · PASCAL-Person-Part | mean IoU (%): 67.89 -> 68.34 | Improved |
| add Inter-Graph Transfer via handcraft relation (ablation) | Graphonomy with CIHP pre-training · PASCAL-Person-Part | mean IoU (%): 70.47 -> 70.22 | Hurt |
| use feature similarity for Inter-Graph Transfer (ablation) | Graphonomy with CIHP pre-training · PASCAL-Person-Part | mean IoU (%): 70.47 -> 71.05 | Improved |
| use learnable matrix for Inter-Graph Transfer (ablation) | Graphonomy with CIHP pre-training · PASCAL-Person-Part | mean IoU (%): 70.47 -> 70.94 | Improved |
| combine feature and semantic similarity transfer (ablation) | Graphonomy · PASCAL-Person-Part | mean IoU (%): 71.05 -> 71.14 | Improved |
| add learnable matrix and handcraft relation to the combination (ablation) | Graphonomy · PASCAL-Person-Part | mean IoU (%): 71.14 -> 70.69 | Hurt |
| replace fine-tuning with Inter-Graph Transfer, 100% data (ablation) | Graphonomy (CIHP) · PASCAL-Person-Part | mean IoU (%): 70.33 -> 71.14 | Improved |
| replace fine-tuning with Inter-Graph Transfer, 50% data (ablation) | Graphonomy (CIHP) · PASCAL-Person-Part | mean IoU (%): 68.45 -> 70.03 | Improved |
| increase Intra-Graph layers from 1 to 3 (ablation) | Intra-Graph Reasoning · PASCAL-Person-Part | mean IoU (%): 67.80 -> 68.34 | Improved |
| increase Intra-Graph layers from 3 to 5 (ablation) | Intra-Graph Reasoning · PASCAL-Person-Part | mean IoU (%): 68.34 -> 68.36 | No clear effect |
| train one universal model over all datasets instead of a dataset-specific transfer | Graphonomy · PASCAL-Person-Part | mean IoU (%): 71.14 -> 70.99 | Hurt |
Closest papers by meaning
- Graphonomy: Universal Human Parsing via Graph Transfer Learning (CVPR 2019 · similarity 0.95)
- Learning Semantic-Specific Graph Representation for Multi-Label Image Recognition (ICCV 2019 · similarity 0.89)
- Semantic Object Parsing with Graph LSTM (ECCV 2016 · similarity 0.88)
- SSTAG: Structure-Aware Self-Supervised Learning Method for Text-Attributed Graphs (NeurIPS 2025 · similarity 0.86)
- SCENIR: Visual Semantic Clarity through Unsupervised Scene Graph Retrieval (ICML 2025 · similarity 0.85)
- GraphVLM: Benchmarking Vision Language Models for Multimodal Graph Learning (CVPR 2026 · similarity 0.85)
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.
