AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

Bayesian Graph Convolution LSTM for Skeleton Based Action Recognition

Rui Zhao, Kang Wang, Hui Su, Qiang Ji

ICCV 2019 · Graph learning › Knowledge and temporal graphs · approach: Graph reasoning for visual recognition · 126 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
add local-graph convolution (vs no graph) (ablation)Bayesian GC-LSTM · UTD-MHADaccuracy (%): 85.0 -> 87.2Improved
add global graph only (vs no graph) (ablation)Bayesian GC-LSTM · UTD-MHADaccuracy (%): 85.0 -> 81.5Hurt
combine local and global graphs (joint graph) (ablation)Bayesian GC-LSTM · UTD-MHADaccuracy (%): 87.2 -> 92.1Improved
mean-field graph (no edges) instead of no graph (ablation)Bayesian GC-LSTM · UTD-MHADaccuracy (%): 85.0 -> 82.2Hurt
LSTM instead of vanilla RNN (ablation)Bayesian GC-LSTM · UTD-MHADaccuracy (%): 66.4 -> 92.1Improved
LSTM instead of GRU (ablation)Bayesian GC-LSTM · UTD-MHADaccuracy (%): 89.7 -> 92.1Improved
stack two LSTM layers or use bi-directional LSTM (ablation)Bayesian GC-LSTM · UTD-MHADaccuracy (%): 92.1 -> 92.0 (stacked), 90.8 (bi-directional)No clear effect
Bayesian inference instead of ML point estimate (clean) (ablation)Bayesian GC-LSTM · UTD-MHAD cleanaccuracy (%): 86.2 -> 87.4Improved
Bayesian inference instead of ML (rotation + noise) (ablation)Bayesian GC-LSTM · UTD-MHAD rotation + noiseaccuracy (%): 65.1 -> 82.8Improved
add prior (MAP instead of ML) under perturbation (ablation)Bayesian GC-LSTM · UTD-MHAD rotation onlyaccuracy (%): 62.8 -> 78.1Improved
add adversarial prior to Bayesian inference (clean) (ablation)Bayesian GC-LSTM · UTD-MHAD cleanaccuracy (%): 87.4 -> 92.1Improved
add adversarial prior to Bayesian inference (rotation + noise) (ablation)Bayesian GC-LSTM · UTD-MHAD rotation + noiseaccuracy (%): 82.8 -> 85.9Improved

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