Research map / Detection and segmentation / Weakly supervised detection and segmentation
Weakly and Semi Supervised Human Body Part Parsing via Pose-Guided Knowledge Transfer
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 |
|---|---|---|---|
| Add keypoint-labeled training data, increasing the keypoint annotations from 0 to 4K | VGG-16 parsing network with 7K mask annotations · PASCAL-Person-Part validation | mIoU: 56.89 -> 59.60 | Helped |
| Add keypoint-labeled training data, increasing the keypoint annotations from 4K to 7K | VGG-16 parsing network with 7K mask annotations · PASCAL-Person-Part validation | mIoU: 59.60 -> 61.44 | Helped |
| Add keypoint-labeled training data, increasing the keypoint annotations from 7K to 10k | VGG-16 parsing network with 7K mask annotations · PASCAL-Person-Part validation | mIoU: 61.44 -> 62.60 | Helped |
| Use pose-guided part-level priors instead of skeleton label-map priors (ablation) | PASCAL-Person-Part parsing network · PASCAL-Person-Part | mIoU: 58.26 -> 62.60 | Helped |
| Use three pose neighbors rather than one to form the part-level prior (ablation) | PASCAL-Person-Part parsing network · PASCAL-Person-Part | mIoU: 60.10 -> 61.78* | Helped |
| Use five rather than three pose neighbors without augmentation (ablation) | PASCAL-Person-Part parsing network · PASCAL-Person-Part | mIoU: 61.78* -> 61.26 | Hurt |
| Use seven rather than three pose neighbors without augmentation (ablation) | PASCAL-Person-Part parsing network · PASCAL-Person-Part | mIoU: 61.78* -> 60.32 | Hurt |
| Add refinement-training augmentation by sampling three candidates from a pool of five (ablation) | Part-level-prior parsing network · PASCAL-Person-Part | mIoU: 61.78* -> 62.60 | Helped |
| Use estimated instead of ground-truth keypoints to generate training labels (ablation) | Semi-supervised parsing network · PASCAL-Person-Part | mIoU: 61.8 mIOU | No clear effect |
| Add keypoint-guided semi-supervised training to the Attention baseline | Horse-Cow baseline · Horse | mIoU: 71.55 -> 74.69 | Helped |
| Add keypoint-guided semi-supervised training to the Attention baseline | Horse-Cow baseline · Cow | mIoU: 68.84 -> 72.23 | Helped |
Closest papers by meaning
- Pose-Guided Human Parsing by an AND/OR Graph Using Pose-Context Features (AAAI 2016 · similarity 0.90)
- Joint Multi-Person Pose Estimation and Semantic Part Segmentation (CVPR 2017 · 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)
- Semantic Human Parsing via Scalable Semantic Transfer Over Multiple Label Domains (CVPR 2023 · similarity 0.88)
- HumanCrafter: Synergizing Generalizable Human Reconstruction and Semantic 3D Segmentation (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.