AI Paper Atlas

Research map / Detection and segmentation / CNN features

DETR Does Not Need Multi-Scale or Locality Design

Yutong Lin, Yuhui Yuan, Zheng Zhang, Chen Li, Nanning Zheng, Han Hu

ICCV 2023 · Detection and segmentation › CNN features · approach: Context for hard detection conditions · 20 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 BoxRPB to cross-attention (ablation)plain DETR (Swin-S) · COCO valAP: 37.2 -> 46.1Improved
add MIM (SimMIM) backbone pre-training (ablation)plain DETR (Swin-S) · COCO valAP: 37.2 -> 44.6Improved
add MIM pre-training on top of BoxRPB (ablation)plain DETR (Swin-S) · COCO valAP: 46.1 -> 48.7Improved
add box re-parameterization for regression (ablation)plain DETR with BoxRPB and MIM · COCO valAP: 48.7 -> 50.9Improved
replace cross-entropy with focal loss for classification (ablation)plain DETR baseline · COCO valAP: 23.1 -> 31.6Improved
add hybrid matching (ablation)plain DETR baseline · COCO valAP: 35.2 -> 37.2Improved
use axial decomposition of BoxRPB (ablation)BoxRPB · COCO valAP and cost: AP 50.8 -> 50.9, GFLOPs 265.4 -> 5.8Improved
use two box corners instead of box center only for BoxRPB (ablation)BoxRPB · COCO valAP: 48.0 -> 50.9Improved
use MIM pre-training to drop multi-scale backbone features (C5 only) (ablation)plain DETR · COCO valAP: supervised: 47.0 -> 46.4; MIM: 50.3 -> 50.2Improved
use multi-scale decoder feature maps instead of single-scale P4 (ablation)plain DETR with MIM · COCO valAP: 50.9 -> 51.1No clear effect
use global BoxRPB cross-attention instead of local cross-attention (deformable, RoIAlign, RoI sampling, box mask) (ablation)plain DETR · COCO valAP: 50.2 (deformable) -> 50.9Improved
replace DAB cross-attention with BoxRPB in DAB official code (ablation)DAB-DETR · COCO valmAP: +1.8 mAPImproved

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