ArcVis

This is a visualization library for 3d/2d plot. It is helpful for developing geometric algorithms and check their correctness.

We use numpy array for 3d/2d data structure, and many of the geometric function are in torch.

We provide some basic geometric implementation and visual samples. For more example usage, please visit our repo ArcNerf. There will be more useful examples using this vis lib.

plot3d

We provide a draw_3d_components that is helpful to plot camera/points/rays/lines/sphere/meshes/volume with customized color in all.

For all point, it should be in world space, or transformed to world space. Then change to plt space. The visual space is (x-right, y-down, z-forward). But we show the y values as -y inorder to make y goes downward.

Backend

We provide matplotlib and plotly for visualization. By default it use plt.

Plotly is good for interactive visualization and is able to zoom-in/out, move. We can add plotly=True in draw_3d_components.

Camera

c2w pose: in (N_cam, 4, 4). Local camera models will be created and transformed into world space
cam_colors: in (N_cam, 3) or (3,) can be used to color each or all cams
intrinsic: in (3, 3) is used to adjust local cam model for actual ray direction.

Points

point: in (N_p, 3) in world space
point_size: is single value used to set each point size
point_color: in (N_p, 3) or (3,) can be used to color each or all points

Lines

lines: a list of lines in (N_pts_in_line, 3) in world space, total N_line, each line prints N_pt_in_line-1 line seg
line_width: is single value or list of value used to set each line width.
line_colors: in (N_line, 3) or (3,) can be used to color each or all lines

Meshes

meshes: a list of mesh of (N_tri, 3, 3) in world space, len is N_m
mesh_colors: color in (N_m, 3) or (3,), applied for each or all mesh
face_colors: color in (N_tri, 3), len is N_m, for each mesh if set.

Rays

rays: a tuple (rays_o, rays_d), each in (N_r, 3), in world coord. rays_d is with actual len, if you want longer arrow, you need to extend rays_d
ray_colors: color in (N_r, 3) or (3,), applied for each or all rays
ray_linewidth: width of ray line, by default is 1

Sphere

sphere_radius: draw a sphere with such radius if not None
sphere_origin: the origin of sphere, by default is (0, 0, 0)

Volume

A complete volume implementation is in geometry.volume. You only input points/lines/faces for visual results.

'grid_pts': grid point, ((n+1)^3, 3). If use corner only, (8,3)
- grid_pts_colors: pts_colors in str. If not exist, use chocolate.
- 'grid_pts_size': pts_size. If not exist, set 20.
'volume_pts': volume point, (n^3, 3)
- 'volume_pts_colors: pts_colors in str. If not exist, use green.
- 'volume_pts_size': pts_size. If not exist, set 20.
'lines': lines of bounding lines or dense lines. list of lines in (2, 3)
'faces': faces of bounding faces or dense faces, np in ((n+1)n^2, 4, 3). regroup each face to 2 triangles for visual
- face_colors: face_colors in str. If not exist, use silver.

Other:

title: title of the fig
save_path: if None, will show the fig, otherwise save it
axis range: update by the component with max values (xyz), show in a cube with same lengths.
return_fig: If set to True and save_path is None, will return the fig(plt) or numpy array(plotly) for further usage.
show_axis: If False, do not show axis but only the fig. By default True.

plot2d

In plot_2d.py, we provide a draw_2d_components that is helpful to plot points/lines in defined colors with legends. We use matplotlib as backend.

Notebook:

We provide the python notebook in draw_3d_examples.ipynb to use the 3d render and draw_2d_examples.ipynb to use the 2d drawer.

TencentARC/ArcVis