Table 6 Qualitative comparison of pose estimation results across different models. The ✓ mark denotes successful estimation (anatomically valid and visually accurate), and ✗ denotes obvious failures.
From: Deep learning-based approaches for human pose estimation in interdisciplinary physics applications
Scene type | OpenPose | AlphaPose | HRNet | HSTPN+APRS (Ours) |
|---|---|---|---|---|
Partial Occlusion (Upper Body) | ✗ | ✓ | ✓ | ✓ |
Crowded Scene (3+ Persons) | ✗ | ✗ | ✓ | ✓ |
Fast Movement (SportsPose) | ✓ | ✗ | ✓ | ✓ |
Low-light Environment | ✗ | ✗ | ✓ | ✓ |
Failure Case 1 (Overlapping Legs) | ✗ | ✗ | ✗ | ✗ |
Failure Case 2 (Full-body Occlusion) | ✗ | ✗ | ✗ | ✗ |
