Deep-Learning-Based Recovery of Missing Optical Marker Trajectories in 3D Motion Capture Systems

As the gap length increased from 1 s to 9 s and the number of gaps from 1 to 5, wand length error showed an increasing trend in all methods, but the proposed U-Bi-LSTM methods demonstrated a smaller increase in wand length error as the gap length and number of gaps increased compared to the state-of-the-art deep learning methods (BAN and A-LSTM+LS).

also in 63,971 other papers