LibVista

Previously libzhifan, libvista is a collection of utility tools for 3D computer vision.

Aims to provide easy-to-use 3D visualization and rendering functionalities for research prototyping.

Example:

from libvista.geometry import projection, SimpleMesh, CameraManager

global_cam = CameraManager(
    fx=5000, fy=5000, cx=228, cy=128,
    img_h=256, img_w=456, in_ndc=False)

mesh = trimesh.load_mesh(mesh_path)
mesh = SimpleMesh(mesh.vertices, mesh.faces, tex_color='purple')

img_pil = Image.open(img_path)

rend = projection.perspective_projection_by_camera(
    mesh, global_cam,
    method=dict(name='pyrender', coor_sys='nr', in_ndc=False),
    image=np.asarray(img_pil))

Image.fromarray(rend).save(pyrender_output_path)

Example output:

Doc

Utility tools.

modules:

• io: txt json pickle reading/writing
• odlib: Visualization for object detection bounding boxes
• epylab: matplotlib extension
• geometry: 3D geometries, camera projections, mesh visualization.
  • The default renderer backend is pyrender.
  • pytorch3d is also supported if installed.
  • The develelopment of this module started with pytorch3d in libzhifan; but libvista makes pytorch3d an optional backend, since pytorch3d is hard to install e.g. on Isambard or Mac OS.
• numeric: Utilities for numeric libraries (numpy, pytorch)
• cvtools: video / frame header and subtitles. Mask overlay.

Unit Tests

pytest

Note that currently pytest still requires pytorch3d to be installed.

Install

Best practice is to specify a commit-hash or tag:

pip install git+https://github.com/zhifanzhu/libvista.git@v0.2.0-2025