AI Paper Atlas

Research map / Detection and segmentation / CNN features

An Economic Framework for 6-DoF Grasp Detection

Xiao-Ming Wu, Jia-Feng Cai, Jian-Jian Jiang, Dian Zheng, Yi-Lin Wei, Wei-Shi Zheng

ECCV 2024 · Detection and segmentation › CNN features · approach: Efficient DETR query design · 34 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
keep all grasp views per point (economic supervision) instead of one grasp per point (ablation)vanilla sparse grasp framework · GraspNet-1Billion Kinectmean AP: 30.34 -> 42.31Improved
add interactive grasp head (ablation)EconomicGrasp · GraspNet-1Billion Kinectmean AP: 42.31 -> 44.14Improved
add composite score estimation alone (without interactive head) (ablation)EconomicGrasp · GraspNet-1Billion Kinectmean AP: 42.31 -> 42.42No clear effect
add composite score estimation on top of interactive head (ablation)EconomicGrasp · GraspNet-1Billion Kinectmean AP: 44.14 -> 44.62Improved
add selective match loss (ablation)EconomicGrasp · GraspNet-1Billion Kinectmean AP: 42.54 -> 44.62Improved
use composite score estimation at inference instead of argmax of classified score (ablation)EconomicGrasp · GraspNet-1Billion Kinectmean AP: 34.98 -> 44.62Improved
add global interactive attention then local interactive attention to the head (ablation)EconomicGrasp · GraspNet-1Billion Kinectmean AP: 42.42 -> 43.49 -> 44.62Improved
modernize S4G sparse baseline with region grouping (ablation)S4G-style sparse baseline · GraspNet-1Billion Kinect seenAP: 18.72 -> 37.38Improved
replace PointNet++ with 3D convolution backbone in vanilla framework (ablation)vanilla sparse framework · GraspNet-1Billion Kinect seenAP: 39.63 -> 43.59Improved
replace dense supervision (GSNet) with economic supervision frameworkGSNet · GraspNet-1Billion Kinect training timetraining time (h): 37.8 -> 8.3Improved
replace GSNet with EconomicGrasp in real-world pick and placeGSNet · Real-world pick and placesuccess rate: 87.8% -> 92.3%Improved
replace GSNet with EconomicGrasp on bad-case countGSNet · GraspNet-1Billion seen scenesscenes with no successful grasp (of 7680): 109/7680 -> 36/7680Improved

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