Table 5.
Recent works on human motion analysis.
| Work | Improved Task | Key Contribution |
|---|---|---|
| [81] | Motion Features Extraction | Spatiotemporal self-similarity |
| [82] | Motion Features Extraction | MotionSqueeze module |
| [83] | Pose Estimation (Key points Positioning) | Multi-branch regression |
| [84] | Pose Estimation (Key points Positioning) | Cascade Transformers |
| [85] | Pose Estimation (Key points Positioning) | Adversarial algorithms |
| [86] | Human Completion | Topological Structure/Memory Bank |
| [87] | Skeleton-Based Action Recognition | Memory Attention Networks |
| [90] | Action Recognition | Temporal-Spatial pooling block |
| [91] | Action Recognition | CNN+Autoencoder+LSTM |
| [101] | Action Recognition | Contrastive Learning |
| [100] | Action Recognition | Semi-supervised Action Detection |
| [95,96,97,98,99] | Action Classification | Transformers |
| [103] | Action Quality Assessment | Contrastive Regression |
| [104] | video representation | Space-Time Graph |
| [105,106] | Video Representation | Self-supervised learning |
| [107] | Temporal Modeling | Two-level Motion Modeling |
| [108] | Motion Segment Extraction | Hierarchical Framework |
| [109] | Temporal Action Localization | E2E anchor free method |
| [110] | Temporal Action Localization | Anchor-Constrained Viterbi |
| [111] | Temporal Action Localization | Memory Network |
| [112] | Temporal Action Localization | Multi-Label Action Dependency layer |
| [113] | Human Object Interaction | Transformer /Cascade detector |
| [114] | Human Object Interaction | Graph Networks |