AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

PromptDet: Towards Open-Vocabulary Detection Using Uncurated Images

Chengjian Feng, Yujie Zhong, Zequn Jie, Xiangxiang Chu, Haibing Ren, Xiaolin Wei, Weidi Xie, Lin Ma

ECCV 2022 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary object detection · 198 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 description to manual prompt (ablation)PromptDet detector · LVIS minivalAP_novel: 7.4 -> 9.0Improved
use regional prompt learning instead of manual description prompt (ablation)PromptDet detector · LVIS minivalAP_novel: 9.0 -> 11.1Improved
add self-training to manual description prompt (ablation)PromptDet · LVIS minivalAP_novel: 9.0 -> 15.3Improved
add self-training to learned prompt (1-iter) (ablation)PromptDet · LVIS minivalAP_novel: 11.1 -> 15.9Improved
add second iteration of prompt learning and image sourcing (ablation)PromptDet · LVIS minivalAP_novel: 15.9 -> 19.0Improved
add third iteration of prompt learning (ablation)PromptDet · LVIS minivalAP_novel: 19.0 -> 19.3No clear effect
use max-prediction-score boxes instead of max-size (ablation)PromptDet · LVIS minivalAP_novel: 9.5 -> 19.0Improved
increase sourced web images from 50 to 300 (ablation)PromptDet · LVIS minivalAP_novel: 14.6 -> 19.0Improved
update RPN classifier and box head during self-training (ablation)PromptDet · LVIS minivalAP_novel: 17.2 -> 19.0Improved
use top-20 proposals instead of all proposals for pseudo labels (ablation)PromptDet · LVIS minivalAP_novel: 10.4 -> 19.0Improved
use top-20 instead of top-30 proposals (ablation)PromptDet · LVIS minivalAP_novel: 16.1 -> 19.0Improved
train longer (12 to 72 epochs)PromptDet · LVIS minivalAP_c: 18.5 -> 23.3Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

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