AI Paper Atlas

Research map / 3D vision / Human pose estimation

Pose Augmentation: Class-agnostic Object Pose Transformation for Object Recognition

Yunhao Ge, Jiaping Zhao, Laurent Itti

ECCV 2020 · 3D vision › Human pose estimation · approach: Category-level object pose estimation · 7 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 OPT-Net eliminate-add generator instead of StarGANpose transformation · iLab-20M pose transformation (6 yaw poses, foreground)MSE: 502.51 -> 374.76Improved
use OPT-Net instead of StarGANpose transformation · iLab-20M pose transformation (6 yaw poses, foreground)PSNR: 21.95 -> 23.04Improved
add OPT-Net synthesized images for missing posesResNet-18 on P-UB · iLab-20M vehicle recognitionoverall accuracy: 52.26% -> 59.15%Improved
add interpolated extra poses to synthesized balanced set (ablation)ResNet-18 on S-P-B · iLab-20M vehicle recognitionoverall accuracy: 59.2 -> 61.2Improved
use OPT-Net synthesis instead of traditional augmentationResNet-18 on P-UB · iLab-20M vehicle recognitionoverall accuracy: 52.3 -> 59.2Improved
add OPT-Net synthesized poses on pretrained-and-finetuned modelAlexNet on RGB-D P-UB · RGB-D object recognitionoverall accuracy: 99.1 -> 99.7Improved
add synthesized poses when only 1 pose per class is availableAlexNet on confusable RGB-D classes · RGB-D confusable categoriesoverall accuracy: 75.1 -> 83.2Improved
add synthesized poses when 2 poses per class are availableAlexNet on confusable RGB-D classes · RGB-D confusable categoriesoverall accuracy: 90.4 -> 94.2Improved
add synthesized poses when 3 poses per class are availableAlexNet on confusable RGB-D classes · RGB-D confusable categoriesoverall accuracy: 99.3 -> 99.4No clear effect
use OPT-Net instead of traditional augmentation with 1 pose per classAlexNet on confusable RGB-D classes · RGB-D confusable categoriesoverall accuracy: 77.6 -> 83.2Improved
pretrain on OPT-Net synthesized images instead of StarGANResNet-18 vehicle classifier · ImageNet car classaccuracy: 41.84 -> 76.20Improved
pretrain AlexNet on OPT-Net synthesized poses instead of scratchAlexNet · ILSVRC-2010, 10 images per classtop-5 accuracy: 4.32 -> 10.94Improved

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