Plan to implement Mesh Class

Currently, a mesh is expressed as a dictionary object that has numpy arrays of vertices and faces as elements. However, it is more convenient if we can deal with mesh as object. Some properties and operations related to meshes should be combined as a class, like this:

import trimesh
import numpy as np

class Mesh(object):
    def __init__(self, vertices: np.ndarray, faces: np.ndarray):
        self.vertices = vertices
        self.faces = faces
        # other variables... 

    @classmethod
    def load_obj(cls, filepath: str):
        mesh = trimesh.load(filepath)
        vertices = mesh.vertices.view(np.ndarray)
        faces = mesh.faces.view(np.ndarray)
        return cls(vertices, faces)

   # other methods... 

If we implement such a class, we can use mesh objects as the arguments of the function to compute metrics and losses about the mesh (e.g., Chamfer Distance).

Do you have any idea?

Start by reviewing the current dictionary-based mesh representations and the metric and loss APIs mentioned in the issue, including Chamfer Distance. No files or tests are named; done would require an agreed mesh-object design that supports the stated vertices, faces, loading, and metric or loss integration.