ikalevatykh / panda3d_viewer

A simple 3D geometry viewer based on Panda3D

panda3d_viewer

A simple 3D geometry viewer based on Panda3D game engine.

Install

Using pip

pip install panda3d_viewer

From source code

git clone https://github.com/ikalevatykh/panda3d_viewer.git
cd panda3d_viewer
python setup.py install

Examples

A simple scene in a GUI window

from panda3d_viewer import Viewer, ViewerConfig

config = ViewerConfig()
config.set_window_size(320, 240)
config.enable_antialiasing(True, multisamples=4)

with Viewer(window_type='onscreen', window_title='example', config=config) as viewer:

    viewer.append_group('root')
    viewer.append_box('root', 'box_node', size=(1, 1, 1))
    viewer.append_sphere('root', 'sphere_node', radius=0.5)

    viewer.set_material('root', 'box_node', color_rgba=(0.7, 0.1, 0.1, 1))
    viewer.set_material('root', 'sphere_node', color_rgba=(0.1, 0.7, 0.1, 1))

    viewer.move_nodes('root', {
        'box_node': ((0, 0, 0.5), (1, 0, 0, 0)),
        'sphere_node': ((0, 0, 1.5), (1, 0, 0, 0))})

    viewer.reset_camera(pos=(4, 4, 2), look_at=(0, 0, 1))
    viewer.save_screenshot(filename='box_and_sphere.png')

Render an animation offscreen

from math import cos, sin, pi
import imageio # install imageio: pip install imageio

from panda3d_viewer import Viewer, ViewerConfig

config = ViewerConfig()
config.set_window_size(320, 240)
config.enable_antialiasing(True, multisamples=4)
config.enable_shadow(True)
config.show_axes(False)
config.show_grid(False)
config.show_floor(True)

viewer = Viewer(window_type='offscreen', config=config)

viewer.append_group('root')
viewer.append_sphere('root', 'sphere_node', radius=0.5)
viewer.set_material('root', 'sphere_node', color_rgba=(0.1, 0.7, 0.1, 1))

with imageio.get_writer('sphere_anim.gif', mode='I') as writer:
    for i in range(50):
        angle = 2 * pi * i / 50
        x = 4 * cos(angle)
        y = 4 * sin(angle)
        z = 0.5 + 0.5 * abs(sin(angle))

        viewer.move_nodes('root', {'sphere_node': ((0, 0, z), (1, 0, 0, 0))})
        viewer.reset_camera(pos=(x, y, 2), look_at=(0, 0, 1))

        image_rgb = viewer.get_screenshot(requested_format='RGB')
        writer.append_data(image_rgb)

Render a point cloud

import numpy as np
import time

from panda3d_viewer import Viewer, ViewerConfig

with Viewer(show_grid=False) as viewer:
    viewer.reset_camera((10, 10, 15), look_at=(0, 0, 0))
    viewer.append_group('root')
    viewer.append_cloud('root', 'cloud', thickness=4)

    while True:
        vertices = np.random.randn(300000, 3).astype(np.float32)
        colors = np.ones((300000, 4), np.float32)
        colors[:, :3] = np.clip(np.abs(vertices), 0, 3) / 3
        viewer.set_cloud_data('root', 'cloud', vertices, colors)
        time.sleep(0.03)

Robotic examples

Using with Pinocchio

Pinocchio is a library for rigid multi-body dynamics computation. To see how to use this package with Pinocchio see example 1, example 2.

Visualize the point cloud from a RealSense camera

import numpy as np
import pyrealsense2 as rs
from panda3d_viewer import Viewer, ViewerConfig

config = rs.config()
config.enable_stream(rs.stream.depth, 640, 480, rs.format.z16, 30)
config.enable_stream(rs.stream.color, 640, 480, rs.format.bgr8, 30)

pipeline = rs.pipeline()
pipeline.start(config)
pc = rs.pointcloud()

camera_frame = (0, 0, 0.2), (0.7071, -0.7071, 0, 0)

with Viewer(window_title='RealSense') as viewer:
    viewer.append_group('root', scale=2)
    viewer.append_cloud('root', 'camera', thickness=4, frame=camera_frame)

    while True:
        frames = pipeline.wait_for_frames()
        color = frames.get_color_frame()
        depth = frames.get_depth_frame()
        if color and depth:
            pc.map_to(color)
            points = pc.calculate(depth)
            vertices = np.asarray(points.get_vertices())
            texture_coords = np.asanyarray(points.get_texture_coordinates())
            texture_image = np.asanyarray(color.get_data())

            viewer.set_cloud_data(
                'root', 'camera',
                vertices=vertices,
                texture_coords=texture_coords,
                texture_image=texture_image)

API

Viewer

This package contains Viewer, a thin wrapper on top of Panda3D ShowBase, an application framework responsible for opening a graphical display, setting up input devices and creating the scene graph.

Viewer constructor takes window_type parameter wich should be one of 'onscreen', 'offscreen'. When selected 'onscreen' the viewer open a GUI window. When selected 'offscreen' the viewer render to an offscreen buffer.

Optional config parameter allows manage the viewer appearance. Configuration provided by ViewerConfig class, which contains methods:

• set_window_size set window size (default: 800x600)
• set_window_fixed disable window resizing (default: on)
• enable_antialiasing turn antialiasing on or off and specify number of MSAA multisamples: 2,4,8,16 (default: off)
• enable_lights turn lighting on or off (default: on)
• enable_shadow turn shadows rendering on or off (default: off)
• enable_hdr turn HDR effect on or off (default: off)
• enable_fog turn fog rendering on or off (default: off)
• show_axes turn the axes rendering on or off (default: on)
• show_grid turn the grid rendering on or off (default: on)
• show_floor turn the floor plane rendering on or off (default: off)

To stop the viewer use stop method. Use join method to wait until a user close the window.

Managing 3D scene

All scene geometries organized in named groups. Each group contains arbitrary number of nodes. To manage the groups use append_group and remove_group. To hide or show all geometries inside a group use show_group.

To append a mesh node to a group use append_mesh, to append a primitive geometry node use append_capsule, append_cylinder, append_box, append_plane, append_sphere. To modify material of a node use set_material.

To specify nodes position in the space use move_nodes function. It takes a dictionary with node_name - position, quaternion pairs, so you can specify the position of all/any nodes in a group simultaneously.

Render to file or memory buffer

To capture an image and save it on the disk use save_screenshot. Specify path to the image file with extention in a filename parameter. If the filename is ommited it will be generated automatically based on the current date time.

To capture an image to a memory buffer use get_screenshot. Specify the color channels order in requested_format parameter. Default format is BGRA, allow any combinations of R,G,B,A channels. The function returns an image as a numpy array.

Scene appearance

To move the camera use reset_camera method which takes the desired camera position and target point.

Viewer provide several methods to manage visual appearance:

• set_background_color change background color
• enable_lights turn lighting on or off
• enable_shadow turn shadows rendering on or off
• enable_hdr turn HDR effect on or off
• enable_fog turn fog rendering on or off
• show_axes turn the axes rendering on or off
• show_grid turn the grid rendering on or off
• show_floor turn the floor plane rendering on or off

Window control

Keyboard shortcuts:

• Show help: F1, h
• Quit window: Escape, q
• Screenshot: Space
• Toggle axes: a
• Toggle HDR: d
• Toggle grid: g
• Toggle fps meter: f
• Toggle lighting: l
• Toggle fog: o
• Toggle plane: p
• Reset camera: r
• Toggle shadows: s
• Toggle texture: t
• Toggle wireframe: w

Mouse control:

• Move: LMB
• Scale: RMB, Ctrl+LMB
• Rotate: LMB+RMB, Alt+LMB
• Tilt: Alt+Ctrl+LMB

License

panda3d_viewer is licensed under the MIT License - see the LICENSE file for details

panda3d_viewer's Issues

Viewer blocks on OSX Catalina

The RealSense example from the README page is blocking on:

ViewerAppProxy(mp.Process):25
reply = self._host_conn.recv()

It does not block if the Viewer is set to use an offline renderer.

macOS Catalina,
Python 3.7
Panda3d via pip (also happens via git install)

Some, but not all, other examples, block too, let me know if you would like me to expand on that.

Rendering artefacts

Hi, your interface with panda3d for robotics applications is promising!

I'm giving it a try but unfortunately the rendering is not as great as I would expect. As you can see on attached screenshot, there is annoying small artefacts on the body the robot. I tried with other robots, and sometimes it is even worse. On the contrary, some other rendering libraries do not produce such artefacts for the same models. Do you have any idea what I'm doing wrong ?

__getstate__ trouble

Windows 10
Python 3.10
Panda3d 1.10.11

Issues when the name argument of ViewerAppProxy.getattr is either getstate or setstate.
(I'm guessing deadlock is occurring. )

I solved it by throwing an exception for these names.

    def __getattr__(self, name):
        """Redirect method calls to the sub-process.

        Arguments:
            name {str} -- attribute name

        Returns:
            callable -- an application method wrapper
        """
        if name.startswith("__"):
            return self.__getattribute__(name)  # raise AttributeError

        def _send(*args, **kwargs):
            self._host_conn.send((name, args, kwargs))
            reply = self._host_conn.recv()
            if isinstance(reply, Exception):
                raise reply
            return reply

        return _send

Thanks for reading.