Fig. 5 | Scientific Reports

Fig. 5

From: Enhanced spatiotemporal skeleton modeling: integrating part-joint attention with dynamic graph convolution

Fig. 5

MPJPE trends on Human3.6M for Walking, Eating, and Smoking at different prediction horizons (80 ms, 160 ms, 320 ms, 400 ms). Our proposed method consistently achieves lower prediction errors across all time steps compared to Res. Sup., convSeq2Seq, and LTD variants, demonstrating its superior capability in capturing fine-grained temporal dynamics and spatial dependencies.

Back to article page