PoseC3D multiple-person scenario

[MABatin]

This is for mmaction2 0.24.1
Hello, I found when exploring the code for demo_video_structuralize.py script, in skeleton_based_stdet:

for i in range(proposal.shape[0]):  # num_person
skeleton_prediction.append([])

This suggests that for multiple-person scenario, the script basically iterates over the pose results of each person and perform predictions individually. However, in the PosC3D paper's abstract I found,
"Also, PoseConv3D can handle multiple-person scenarios without additional computation costs."
Is there a way to perform predictions for multiple-persons without the need for explicit for loop. If so, then how?