Table 1 Accuracy comparisons between our proposed STG-NODE model and other methods on NTU RGB+D 60 and Kinetics Skeleton 400. Significant values are in bold.
From: Spatial-temporal graph neural ODE networks for skeleton-based action recognition
Method | X-Sub | X-View | Kinetics Top1 | Kinetics Top5 | Years |
|---|---|---|---|---|---|
Deep LSTM13 | 60.7 | 67.3 | 16.4 | 35.3 | 2016 |
TCN11 | 74.3 | 83.1 | 20.3 | 40.0 | 2017 |
ST-GCN27 | 81.5 | 88.3 | 30.7 | 52.8 | 2018 |
DS-LSTN43 | 75.5 | 84.2 | - | - | 2020 |
STD+RGB-DI44 | 79.4 | 84.1 | - | - | 2020 |
GFNet45 | 82.0 | 89.9 | - | - | 2020 |
STA46 | 72.4 | 79.7 | - | - | 2021 |
CNN+LSTM47 | 81.9 | 88.7 | - | - | 2021 |
PoT2I48 | 83.9 | 90.3 | - | - | 2021 |
C-CNN+HTLN49 | 83.5 | 86.8 | - | - | 2022 |
Custom ST-GCN41 | 82.7 | 90.2 | 32.3 | 54.5 | 2023 |
STG-NODE (ours) | 84.0 | 91.1 | 32.6 | 55.0 | 2023 |
