Research map / Detection and segmentation / CNN features
Parsing R-CNN for Instance-Level Human Analysis
Lu Yang, Qing Song, Zhihui Wang, Ming Jiang
Check 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 |
|---|---|---|---|
| Use P2 features only for parsing while keeping pyramid scale assignment for detection (ablation) | Parsing R-CNN · CIHP val | mIoU: 47.2 -> 48.2 | Improved |
| Use P2 only for both bbox and parsing branches (ablation) | Parsing R-CNN · CIHP val | mIoU: 48.2 -> 47.7 | Hurt |
| Increase parsing RoI resolution from 14×14 to 32×32 (ablation) | Parsing R-CNN · CIHP val | mIoU: 48.2 -> 50.7 | Improved |
| Use 100 RoIs at inference with 32×32 parsing RoIs (ablation) | Parsing R-CNN · CIHP val | mIoU: 50.7 -> 50.5 | No clear effect |
| Increase parsing RoI resolution to 64×64 (ablation) | Parsing R-CNN · CIHP val | mIoU: 50.7 -> 51.5 | Improved |
| Use ASPP only instead of the baseline parsing branch (ablation) | Parsing R-CNN · CIHP val | mIoU: 50.7 -> 51.9 | Improved |
| Use Non-local only without ASPP (ablation) | Parsing R-CNN · CIHP val | mIoU: 50.7 -> 50.5 | Hurt |
| Add combined ASPP and Non-local GCE (ablation) | Parsing R-CNN · CIHP val | mIoU: 50.7 -> 52.7 | Improved |
| Add four convolutional layers before GCE (ablation) | Parsing R-CNN · CIHP val | mIoU: 52.7 -> 52.8 | No clear effect |
| Add four convolutional layers after GCE (ablation) | Parsing R-CNN · CIHP val | mIoU: 52.7 -> 53.5 | Improved |
| Add four convolutional layers both before and after GCE (ablation) | Parsing R-CNN · CIHP val | mIoU: 52.7 -> 53.1 | Improved |
| Extend training from 1x to 3x iterations (ablation) | Parsing R-CNN · CIHP val | mIoU: 53.5 -> 56.3 | Improved |
Closest papers by meaning
- Instance-level Human Parsing via Part Grouping Network (ECCV 2018 · similarity 0.90)
- Renovating Parsing R-CNN for Accurate Multiple Human Parsing (ECCV 2020 · similarity 0.90)
- Deep Learning Technique for Human Parsing: A Survey and Outlook (IJCV 2024 · similarity 0.87)
- Differentiable Multi-Granularity Human Parsing (TPAMI 2023 · similarity 0.87)
- Differentiable Multi-Granularity Human Representation Learning for Instance-Aware Human Semantic Parsing (CVPR 2021 · similarity 0.86)
- Look Into Person: Self-Supervised Structure-Sensitive Learning and a New Benchmark for Human Parsing (CVPR 2017 · similarity 0.86)
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.
