Visualizing a sequence of point clouds in Open3D (0.10.0) as a video

Question

Suppose we have a sequence of .pcd files in the directory point_clouds:

01.pcd
02.pcd
...
10.pcd

I am trying to visualize those files as a video (and potentially save them as a *mp4/avi). I tried two methods:

1. Inspired by the Non-blocking visualization guide, this is the code I made:

import open3d as o3d
from pathlib import Path

pcs_path = Path("point_clouds")
pcd = o3d.io.read_point_cloud(str(pcs_path.joinpath("01.pcd")))

vis = o3d.visualization.Visualizer()
vis.create_window()
vis.add_geometry(pcd)

for p in pcs_path.iterdir():
    pcd = o3d.io.read_point_cloud(str(p))
    
    vis.update_geometry(pcd)
    vis.poll_events()
    vis.update_renderer()

vis.destroy_window()

The issue is that the video is stuck on the very first point cloud, without changing. I was wondering if I should update the pcd "inplace".

2. I used the draw_geometries_with_custom_animation(...) method:

list_of_pcs = [o3d.io.read_point_cloud(str(p)) for p in pcs_path.iterdir()]
o3d.visualization.draw_geometries_with_custom_animation(list_of_pcs)

The issue here is that instead of a sequence, all point clouds are shown simultaneously.

Answer 1

I couldn't get update_geometry to work, even following code that worked perfectly for others. My work around was to add a new geometry each time, which lets you get an animation, although it is a bit jittery:

import open3d as o3d
from natsort import natsorted
import os
import time

input_dir = "C:\Users\me\imagedir"
input_files = natsorted(os.listdir(input_dir))
    
# read first ply file
pcd = o3d.io.read_point_cloud(os.path.join(input_dir, input_files[0]))
vis_transform = [[1,0,0,0],[0,-1,0,0],[0,0,-1,0],[0,0,0,1]] # transforms to make viewpoint match camera perspective
pcd.transform(vis_transform)

# getting open3d to display the video
vis = o3d.visualization.Visualizer()
vis.create_window()

# iterate through remaining files     
for input_file in input_files[1:]:
    input_file = os.path.join(input_dir, input_file)
    
    # remove previous pointcloud and add new one    
    vis.remove_geometry(pcd,False)
    pcd = o3d.io.read_point_cloud(input_file) # Read the point cloud
    pcd.transform(vis_transform)
        
    # add pcd to visualizer
    vis.add_geometry(pcd)
    vis.poll_events()
    vis.update_renderer()
    time.sleep(0.005) # makes display a little smoother (display will not match fps of video)

vis.destroy_window()