AI Paper Atlas

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

CoTDet: Affordance Knowledge Prompting for Task Driven Object Detection

Jiajin Tang, Ge Zheng, Jingyi Yu, Sibei Yang

ICCV 2023 · Detection and segmentation › Open-set and few-shot detection · approach: Object-level representation learning · 32 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
use affordance-level visual knowledge instead of object categories only (ablation)CoTDet · COCO-TasksmAP_box: 31.9 -> 48.1Improved
generate visual features per retrieved object instead of one direct query (ablation)CoTDet · COCO-TasksmAP_box: 48.1 -> 52.4Improved
add rationales to MLCoT (ablation)CoTDet · COCO-TasksmAP_box: 52.4 -> 56.9Improved
add rationales to MLCoT (ablation)CoTDet · COCO-TasksmAP_mask: 49.0 -> 53.6Improved
use ChatGPT instead of GPT-3 for knowledge (ablation)CoTDet · COCO-TasksmAP_box: 56.9 -> 57.0No clear effect
use knowledge to initialize queries instead of Def+TOIST text queries (ablation)Deformable-DETR · COCO-TasksmAP_box: 40.3 -> 48.7Improved
fuse knowledge with the visual feature map (ablation)Init w/ MLCoT · COCO-TasksmAP_box: 48.7 -> 50.6Improved
knowledge-conditional query generation (select) instead of fusion (ablation)Fuse w/ MLCoT · COCO-TasksmAP_box: 50.6 -> 55.3Improved
add knowledge-conditional denoising training (ablation)CoTDet · COCO-TasksmAP_box: 55.3 -> 56.9Improved
add knowledge-conditional denoising training (ablation)CoTDet · COCO-TasksmAP_mask: 51.7 -> 53.6Improved
replace TOIST text-initialized queries with full CoTDet decoder (ablation)Def+TOIST · COCO-TasksmAP_box: 40.3 -> 56.9Improved
replace GGNN/TOIST baselines with CoTDetTOIST · COCO-Tasks detectionmAP_box: 41.3 -> 56.9Improved

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