Research map / Detection and segmentation / Open-set and few-shot detection
PartGLEE: A Foundation Model for Recognizing and Parsing Any Objects
Junyi Li, Junfeng Wu, Weizhi Zhao, Song Bai, Xiang Bai
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 Q-Former to independent decoders so parts are parsed from object queries (ablation) | PartGLEE (Pascal Part Base + VOC) · Zero-shot PartImageNet | segm AP: 5.6 -> 6.9 | Improved |
| Add object-level VOC data to training (ablation) | PartGLEE · Pascal Part novel parts | NovelAP: 2.1 -> 5.8 | Improved |
| Add object-level VOC data to training (ablation) | VLPart · Pascal Part novel parts | NovelAP: 1.5 -> 2.1 | Improved |
| Add parsed ImageNet pseudo-labels to PartGLEE training (ablation) | PartGLEE · PartImageNet cross-dataset | mAP_mask: 9.9 -> 14.9 | Improved |
| Add image-level IN-S20 label data to VLPart (ablation) | VLPart · Pascal Part novel parts | NovelAP: 1.5 -> 1.1 | Hurt |
| Use independent object and part decoders instead of a single decoder (ablation) | PartGLEE · COCO-val, LVIS-minival, PACO | AP: COCO AP_box 47.8 -> 49.3 | Improved |
| Add parallel pixel decoders (ablation) | PartGLEE · joint-training ablation | AP: PartImageNet 39.0 -> 38.3 | No clear effect |
| Increase topK and number of parsing queries L (ablation) | PartGLEE · PACO | AP_mask^opart: 16.8 -> 17.4 | No clear effect |
| Joint-train on object and part data (PartGLEE) vs VLPart | PartGLEE(L) vs VLPart(B) · PACO | AP_mask: 22.1 -> 27.8 | Improved |
| Add part-level training on top of GLEE | GLEE(R50) · COCO-val | AP_mask: 48.4 -> 47.6 | Hurt |
| Use two-stage-free one-stage PartGLEE vs OV-PARTS baselines | ADE-Part-234 Oracle-Obj · ADE-Part-234 | Harmonic IoU: 33.85 -> 41.83 | Improved |
Later papers that compared with this method
2 later accepted papers report PartGLEE as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- LangHOPS: Language Grounded Hierarchical Open-Vocabulary Part Segmentation (NeurIPS 2025 · better on 2 of 2 shared results)
- Fine-Grained Image-Text Correspondence with Cost Aggregation for Open-Vocabulary Part Segmentation (CVPR 2025 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PARTONOMY: Large Multimodal Models with Part-Level Visual Understanding (NeurIPS 2025 · similarity 0.88)
- Fine-Grained Image-Text Correspondence with Cost Aggregation for Open-Vocabulary Part Segmentation (CVPR 2025 · similarity 0.88)
- Multi-Part Object Representations via Graph Structures and Co-Part Discovery (NeurIPS 2026 · similarity 0.87)
- Understanding Multi-Granularity for Open-Vocabulary Part Segmentation (NeurIPS 2024 · similarity 0.87)
- LangHOPS: Language Grounded Hierarchical Open-Vocabulary Part Segmentation (NeurIPS 2025 · similarity 0.86)
- CALICO: Part-Focused Semantic Co-Segmentation with Large Vision-Language Models (CVPR 2025 · 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.
