About RGBD visualization using open3d

[crazyggg1]

Hello! First, thank you for your great work on Fast3R. I'm trying to export camera poses and depth maps for downstream applications, but I'm encountering misalignment issues when visualizing results in Open3D and during view projection.
I'm exporting data using this code:

for view_idx, (view, camera_pose, focal, ptc) in enumerate(zip(output_dict['views'], camera_poses, focals, output_dict['preds'])):
    img_name = f"{view_idx:05d}.png"
    img = view["img"][0]
    true_shape = view["true_shape"][0]
    height, width = true_shape
    cx = width / 2.0
    cy = height / 2.0
    intrinsic = np.array([
        [focal, 0,  cx],
        [0,  focal, cy],
        [0,  0,  1]
    ])
    depthmap = ptc['depth']
    # depthmap = ptc['pts3d_in_other_view'][:, :, :, -1][0].cpu().numpy()
    bgr_img = cv2.cvtColor(rgb(img), cv2.COLOR_RGB2BGR)
    cv2.imwrite(os.path.join(imgs_save_dir, img_name), bgr_img * 255)
    cv2.imwrite(os.path.join(depths_save_dir, img_name), depthmap * 255)
    np.savez(
        os.path.join(camera_poses_save_dir, img_name.replace("png", "npz")),
        intrinsics=intrinsic,
        cam2world=camera_pose,
    )