Research map / 3D vision / Human pose estimation
Modulated Graph Convolutional Network for 3D Human Pose Estimation
Zhiming Zou, Wei Tang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add learnable additive affinity modulation A_add instead of multiplicative mask A_mul (ablation) | Vanilla GCN · Human3.6M Protocol #1 (GT 2D, ablation, 128 ch) | MPJPE: 43.05 -> 40.53 | Improved |
| add skeleton prior to the learnable affinity (A_add vs A_no-skeleton) (ablation) | Vanilla GCN · Human3.6M Protocol #1 (GT 2D, ablation, 128 ch) | MPJPE: 42.21 -> 40.53 | Improved |
| use purely additive A_add instead of mixed A_mix (ablation) | Vanilla GCN · Human3.6M Protocol #1 (GT 2D, ablation, 128 ch) | MPJPE: 41.10 -> 40.53 | Improved |
| make the affinity learnable instead of fixed skeleton graph (ablation) | Vanilla GCN · Human3.6M Protocol #1 (GT 2D, ablation, 128 ch) | MPJPE: 49.73 -> 40.53 | Improved |
| add symmetry regularization to affinity modulation (ablation) | Vanilla GCN + A_add · Human3.6M Protocol #1 (GT 2D, ablation, 128 ch) | MPJPE: 40.53 -> 39.42 | Improved |
| add sparsity regularization to affinity modulation (ablation) | Vanilla GCN + A_add · Human3.6M Protocol #1 (GT 2D, ablation, 128 ch) | MPJPE: 40.53 -> 40.23 | Improved |
| add low-rank regularization to affinity modulation (ablation) | Vanilla GCN + A_add · Human3.6M Protocol #1 (GT 2D, ablation, 128 ch) | MPJPE: 40.53 -> 39.43 | No clear effect |
| add L1 loss on prediction error alongside L2 (ablation) | Vanilla GCN + A_add + symmetry · Human3.6M Protocol #1 (GT 2D, ablation, 128 ch) | MPJPE: 39.42 -> 38.52 | Improved |
| add sparsity on top of symmetry regularization (ablation) | Vanilla GCN + A_add · Human3.6M Protocol #1 (GT 2D, ablation, 128 ch) | MPJPE | Hurt |
| add weight modulation instead of weight sharing (128 ch) (ablation) | Vanilla GCN + A_add · Human3.6M Protocol #1 (GT 2D, ablation) | MPJPE: 40.53 -> 38.83 | Improved |
| add weight modulation instead of weight sharing (256 ch) (ablation) | Vanilla GCN + A_add · Human3.6M Protocol #1 (GT 2D, ablation) | MPJPE: 39.39 -> 37.43 | Improved |
| use weight modulation instead of weight unsharing (256 ch) (ablation) | Vanilla GCN + A_add · Human3.6M Protocol #1 (GT 2D, ablation) | MPJPE: 39.09 -> 37.43 | Improved |
Closest papers by meaning
- A Comprehensive Study of Weight Sharing in Graph Networks for 3D Human Pose Estimation (ECCV 2020 · similarity 0.93)
- Locally Connected Network for Monocular 3D Human Pose Estimation (TPAMI 2020 · similarity 0.91)
- Optimizing Network Structure for 3D Human Pose Estimation (ICCV 2019 · similarity 0.91)
- GLA-GCN: Global-local Adaptive Graph Convolutional Network for 3D Human Pose Estimation from Monocular Video (ICCV 2023 · similarity 0.90)
- Semantic Graph Convolutional Networks for 3D Human Pose Regression (CVPR 2019 · similarity 0.90)
- GCNext: Towards the Unity of Graph Convolutions for Human Motion Prediction (AAAI 2024 · similarity 0.88)
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.
