Research map / Detection and segmentation / Open-set and few-shot detection
"PartImageNet: A Large, High-Quality Dataset of Parts"
Ju He, Shuo Yang, Shaokang Yang, Adam Kortylewski, Xiaoding Yuan, Jie-Neng Chen, Shuai Liu, Cheng Yang, Qihang Yu, Alan Yuille
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 part-annotation deep supervision at stage 2 (ablation) | Deeplab v3+ · PartImageNet object segmentation | mIoU (val): 74.04 -> 74.21 | No clear effect |
| add part-annotation deep supervision at stage 3 (ablation) | Deeplab v3+ · PartImageNet object segmentation | mIoU (test): 71.79 -> 72.26 | Improved |
| add part-annotation deep supervision at stage 4 (ablation) | Deeplab v3+ · PartImageNet object segmentation | mIoU (val): 74.04 -> 75.67 | Improved |
| add part-annotation deep supervision at stage 4 (ablation) | Deeplab v3+ · PartImageNet object segmentation | mIoU (test): 71.79 -> 72.84 | Improved |
| add object masks as deep supervision at stage 4 (ablation) | Deeplab v3+ · PartImageNet object segmentation | mIoU (val): 74.04 -> 72.45 | Hurt |
| supervise parts with cross entropy instead of none (ablation) | Deeplab v3+ stage 4 · PartImageNet object segmentation | mIoU (val): 74.04 -> 73.88 | Hurt |
| use binary cross entropy instead of cross entropy for part deep supervision (ablation) | Deeplab v3+ stage 4 · PartImageNet object segmentation | mIoU (val): 73.88 -> 75.67 | Improved |
| initialize part dictionaries with part annotations (ablation) | COMPAS · PartImageNet few-shot 5-way 1-shot | accuracy: 67.1 -> 68.0 | Improved |
| initialize part dictionaries with part annotations (ablation) | COMPAS · PartImageNet few-shot 5-way 5-shot | accuracy: 82.3 -> 82.9 | Improved |
| use annotated part regions instead of random regions (ablation) | DeepEMD · PartImageNet few-shot 5-way 1-shot | accuracy: 67.3 -> 68.5 | Improved |
| use annotated part regions instead of random regions (ablation) | DeepEMD · PartImageNet few-shot 5-way 5-shot | accuracy: 82.7 -> 83.6 | Improved |
Closest papers by meaning
- PartImageNet++ Dataset: Scaling up Part-based Models for Robust Recognition (ECCV 2024 · similarity 0.91)
- Towards Open-World Segmentation of Parts (CVPR 2023 · similarity 0.91)
- From Pose to Part: Weakly-Supervised Pose Evolution for Human Part Segmentation (TPAMI 2022 · similarity 0.90)
- PARTONOMY: Large Multimodal Models with Part-Level Visual Understanding (NeurIPS 2025 · similarity 0.89)
- Fine-Grained Image-Text Correspondence with Cost Aggregation for Open-Vocabulary Part Segmentation (CVPR 2025 · similarity 0.89)
- Understanding Multi-Granularity for Open-Vocabulary Part Segmentation (NeurIPS 2024 · similarity 0.88)
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.
