Fig 1.
Multi-person dance tiered posture recognition with cross progressive multi-resolution representation integration.
Fig 2.
Human skeleton keypoints model.
(a) 15 key points. (b) 18 key points. (c) 25 key points.
Fig 3.
Multi-person posture recognition process.
(a) Top-down. (b) Bottom-up.
Fig 4.
YOLOv8.
Table 1.
Detailed parameter settings of used YOLOV8.
Fig 5.
HRNET model.
Fig 6.
CPMRI module.
Fig 7.
Geometrically-based keypoint segmentation.
Table 2.
The results of the ablation experiments.
Fig 8.
The training curve of the different ablation models.
Fig 9.
The testing curve of the different ablation models.
Fig 10.
Visualized results in the MSCOCO2017 dataset.
Table 3.
The results of the contarst experiments.
Underlined denotes SOTA top-down approaches, italicized denotes SOTA bottom-up methods, and bold represents the best result in each metric.
Fig 11.
Visualized single-person results in the custom dataset.
Fig 12.
Visualized multi-person results in the custom dataset.