AI Paper Atlas

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

EdaDet: Open-Vocabulary Object Detection Using Early Dense Alignment

Cheng Shi, Sibei Yang

ICCV 2023 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary object detection · 53 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
replace object-level alignment with dense-level alignment (ablation)EdaDet · COCOnovel box AP50: 15.1 -> 30.1Improved
add top-k mask M_k (k=12x12) over box scores (ablation)EdaDet dense alignment · COCOnovel box AP50: 30.1 -> 33.1Improved
use a small top-k mask (7x7) instead of 12x12 (ablation)EdaDet dense alignment · COCOnovel box AP50: 33.1 -> 29.3Hurt
fuse CLIP dense probability map (lambda=0.25) (ablation)EdaDet · COCOnovel box AP50: 33.1 -> 35.7Improved
add global alignment to CLIP image feature (ablation)EdaDet · COCOnovel box AP50: 35.7 -> 37.8Improved
deeply decouple proposal generation and classification branches (ablation)CLIP-classifier detector · COCOnovel box AP50: 7.6 -> 14.2Improved
deeply decoupled DETR-style proposal head instead of Mask-RCNN RPN or Def-DETR (ablation)proposal generation · LVIS base categoriestop-300 AR: 0.33 -> 0.36Improved
add prompt ensembling to the classifier (ablation)decoupled CLIP detector · COCOnovel box AP50: 14.2 -> 15.1Improved
apply dense alignment only at one early stage (conv2_x or conv3_x) instead of both (ablation)EdaDet · COCOnovel box AP50: 37.8 -> 35.9 / 36.2Hurt
add self-training (ablation)EdaDet · COCOnovel box AP50: 37.8 -> 40.2Improved
train longer (10 to 30 epochs) with ViT-BaseEdaDet-ViT-Base · LVISnovel mask AP: 29.9 -> 35.6Improved

Later papers that compared with this method

2 later accepted papers report EdaDet 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