Research map / Label-efficient and robust learning / Recognition and long-tail
Semantic Human Parsing via Scalable Semantic Transfer Over Multiple Label Domains
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 across label domains using SST rather than plain multi-dataset training | Human parsing network with ResNet-50 · PASCAL-Person-Part | mIoU: 70.40 -> 72.25 | Helped |
| Use SST for universal parsing rather than plain multi-dataset training | Human parsing network with ResNet-50 · ATR | mIoU: 79.39 -> 79.97 | Helped |
| Use SST for universal parsing rather than plain multi-dataset training | Human parsing network with ResNet-50 · CIHP | mIoU: 62.74 -> 63.35 | Helped |
| Use full pairwise dataset connections instead of progressive connections | SST · PASCAL, ATR, and CIHP | mIoU: 72.03 -> 72.25; 79.64 -> 79.97; 63.19 -> 63.35 | Helped |
| Transfer with SST rather than direct fine-tuning, using 50% target data | CIHP-pretrained model adapted to PASCAL-Person-Part · PASCAL-Person-Part | mIoU: 69.23 -> 71.81 | Helped |
| Transfer with SST rather than direct fine-tuning, using all target data | CIHP-pretrained model adapted to PASCAL-Person-Part · PASCAL-Person-Part | mIoU: 71.52 -> 73.49 | Helped |
| Add MSE with the intra-domain connectivity prior (ablation) | Multi-dataset parsing model · PASCAL-Person-Part | mIoU: 70.40 -> 71.02 | Helped |
| Add the intra-domain connectivity prior to MSE (ablation) | MSE-enabled multi-dataset parsing model · PASCAL-Person-Part | mIoU: 70.78 -> 71.02 | Helped |
| Add dataset-level semantic consistency regularization (ablation) | MSE-enabled multi-dataset parsing model · PASCAL-Person-Part | mIoU: +0.43% | Helped |
| Add image-level semantic consistency regularization (ablation) | MSE-enabled multi-dataset parsing model · PASCAL-Person-Part | mIoU: +0.91% | Helped |
| Add both dataset-level and image-level semantic consistency regularization (ablation) | MSE-enabled multi-dataset parsing model · PASCAL-Person-Part | mIoU: 71.02 -> 72.25 | Helped |
Closest papers by meaning
- Graphonomy: Universal Human Parsing via Graph Transfer Learning (CVPR 2019 · similarity 0.91)
- Part-Aware Context Network for Human Parsing (CVPR 2020 · similarity 0.89)
- Look Into Person: Self-Supervised Structure-Sensitive Learning and a New Benchmark for Human Parsing (CVPR 2017 · similarity 0.88)
- CDGNet: Class Distribution Guided Network for Human Parsing (CVPR 2022 · similarity 0.88)
- Self-Learning With Rectification Strategy for Human Parsing (CVPR 2020 · similarity 0.88)
- ACT as Human: Multimodal Large Language Model Data Annotation with Critical Thinking (NeurIPS 2025 · 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.