the C_obs obtained after running the following code are all 0

[zhangyu2234]

    Dear author, first of all, thank you for providing such a wonderful work, but I have a question. When processing the ETH dataset, the C_obs obtained after running the following code are all 0. Can you tell me why this is? Is there an error in one of my steps?
    
    
    n_ped = obs_traj.size(0)

    # Filter out static trajectory
    mask = (obs_traj[:, -1] - obs_traj[:, -3]).div(2).norm(p=2, dim=-1) > self.static_dist
    obs_m_traj = obs_traj[mask]
    obs_s_traj = obs_traj[~mask]
    pred_m_traj_gt = pred_traj[mask] if pred_traj is not None else None
    pred_s_traj_gt = pred_traj[~mask] if pred_traj is not None else None

    # Projection
    C_m_obs, C_m_pred_gt = self.ET_m_descriptor.projection(obs_m_traj, pred_m_traj_gt)
    C_s_obs, C_s_pred_gt = self.ET_s_descriptor.projection(obs_s_traj, pred_s_traj_gt)
    C_obs = torch.zeros((self.k, n_ped), dtype=torch.float, device=obs_traj.device)
    C_obs[:, mask], C_obs[:, ~mask] = C_m_obs, C_s_obs  # KN

[InhwanBae]

Hi @zhangyu2234,

Thank you for your interest in my work! I’ve just re-run the code to double-check, and I can confirm that the values are being processed correctly. Could you let me know which baseline model you’re using? In the code, a temporary zero-filled C_obs is created and then filled using C_m_obs and C_s_obs. Could you verify if C_m_obs and C_s_obs are also zero? If you could provide more details about the issue, I’d be happy to help with debugging.