AI Paper Atlas

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

ECCV 2024 · Detection and segmentation › Open-set and few-shot detection · approach: Object-level representation learning · 8 citations (OpenAlex)

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 changedWhereMeasuredEffect
Add Q-Former to independent decoders so parts are parsed from object queries (ablation)PartGLEE (Pascal Part Base + VOC) · Zero-shot PartImageNetsegm AP: 5.6 -> 6.9Improved
Add object-level VOC data to training (ablation)PartGLEE · Pascal Part novel partsNovelAP: 2.1 -> 5.8Improved
Add object-level VOC data to training (ablation)VLPart · Pascal Part novel partsNovelAP: 1.5 -> 2.1Improved
Add parsed ImageNet pseudo-labels to PartGLEE training (ablation)PartGLEE · PartImageNet cross-datasetmAP_mask: 9.9 -> 14.9Improved
Add image-level IN-S20 label data to VLPart (ablation)VLPart · Pascal Part novel partsNovelAP: 1.5 -> 1.1Hurt
Use independent object and part decoders instead of a single decoder (ablation)PartGLEE · COCO-val, LVIS-minival, PACOAP: COCO AP_box 47.8 -> 49.3Improved
Add parallel pixel decoders (ablation)PartGLEE · joint-training ablationAP: PartImageNet 39.0 -> 38.3No clear effect
Increase topK and number of parsing queries L (ablation)PartGLEE · PACOAP_mask^opart: 16.8 -> 17.4No clear effect
Joint-train on object and part data (PartGLEE) vs VLPartPartGLEE(L) vs VLPart(B) · PACOAP_mask: 22.1 -> 27.8Improved
Add part-level training on top of GLEEGLEE(R50) · COCO-valAP_mask: 48.4 -> 47.6Hurt
Use two-stage-free one-stage PartGLEE vs OV-PARTS baselinesADE-Part-234 Oracle-Obj · ADE-Part-234Harmonic IoU: 33.85 -> 41.83Improved

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.

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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy