AI Paper Atlas

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

Sim2Real Object-Centric Keypoint Detection and Description

Chengliang Zhong, Chao Yang, Fuchun Sun, Jinshan Qi, Xiaodong Mu, Huaping Liu, Wenbing Huang

AAAI 2022 · Detection and segmentation › Open-set and few-shot detection · approach: Object-centric learning theory and video · 11 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 object-centric negative sampling instead of random pixel negatives (ablation)Object-centric keypoint learning · Synthetic-real image matchingMMA5: 1.1% -> 46.9%Improved
Add detector repeatability loss to object-centric sampling (ablation)Object-centric keypoint learning · Real-real image matchingMMA5: 29.3% -> 35.1%Improved
Add decoupled descriptors (ablation)Object-centric keypoint learning · Real-real image matchingMMA5: 35.1% -> 44.4%Improved
Add semantic consistency training (ablation)Object-centric keypoint learning · 6D pose estimation on YCB-VideoADD: 40.8 -> 42.2Improved
Set confidence threshold to 0.0 rather than 1.5 (ablation)Object-centric keypoint detector · YCB-Video pose estimationADD: 42.2 -> 15.8Hurt
Use confidence threshold 3.0 rather than 1.5 (ablation)Object-centric keypoint detector · YCB-Video pose estimationADD: 42.2 -> 40.2Hurt
Use confidence threshold 4.5 rather than 1.5 (ablation)Object-centric keypoint detector · YCB-Video pose estimationADD: 42.2 -> 36.1Hurt
Set inter-object InfoNCE temperature to 0.07 rather than 0.2 (ablation)Object-centric descriptor · YCB-Video pose estimationADD: 42.2 -> 40.2Hurt
Set intra-object temperature to 0.2 rather than 0.07 (ablation)Object-centric descriptor · YCB-Video pose estimationADD: 42.2 -> 32.1Hurt
Increase inter-object descriptor dimension from 32 to 64 (ablation)Object-centric descriptor · YCB-Video pose estimationADD: 41.1 -> 39.0Hurt
Increase pose templates from 48 to 96 (ablation)Object-centric pose estimation · YCB-VideoADDS: 55.8 -> 61.9Improved
Use two-stage object-centric matching instead of direct concatenated-descriptor matching (ablation)Object-centric keypoint matching · YCB-Video pose estimationADD: 30.5 -> 42.2Improved

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