How to visualize voxels and show it on meshlab?

Question

I have a voxel coordinates tensor with shape (Number_voxel,4). The second dim just indicate the batch_id and coordinates of that voxel (batch_id,x,y,z).

How can I use this tensor to create voxels which can be seen on meshlab?

Currently, I use these codes, but the result is not good.

 import numpy as np
 import os
 from plyfile import PlyData, PlyElement



def write_ply(points, face_data, filename, text=True):

  points = [(points[i,0], points[i,1], points[i,2]) for i in range(points.shape[0])]

  vertex = np.array(points, dtype=[('x', 'f4'), ('y', 'f4'),('z', 'f4')])

  face = np.empty(len(face_data),dtype=[('vertex_indices', 'i4', (4,))])
  face['vertex_indices'] = face_data

  ply_faces = PlyElement.describe(face, 'face')
  ply_vertexs = PlyElement.describe(vertex, 'vertex', comments=['vertices'])
  PlyData([ply_vertexs, ply_faces], text=text).write(filename)

def occ2points(coordinates):
  points  = []
  len = coordinates.shape[0]
  for i in range(len):
    points.append(np.array([int(coordinates[i,1]),int(coordinates[i,2]),int(coordinates[i,3])]))
 
  return np.array(points)

def generate_faces(points):
  corners = np.zeros((8*len(points),3))
  faces = np.zeros((6*len(points),4))
  for index in range(len(points)):
    corners[index*8]= np.array([points[index,0]-0.5, points[index,1]-0.5, points[index,2]-0.5])
    corners[index*8+1]= np.array([points[index,0]+0.5, points[index,1]-0.5, points[index,2]-0.5])
    corners[index*8+2]= np.array([points[index,0]-0.5, points[index,1]+0.5, points[index,2]-0.5])
    corners[index*8+3]= np.array([points[index,0]+0.5, points[index,1]+0.5, points[index,2]-0.5])
    corners[index*8+4]= np.array([points[index,0]-0.5, points[index,1]-0.5, points[index,2]+0.5])
    corners[index*8+5]= np.array([points[index,0]+0.5, points[index,1]-0.5, points[index,2]+0.5])
    corners[index*8+6]= np.array([points[index,0]-0.5, points[index,1]+0.5, points[index,2]+0.5])
    corners[index*8+7]= np.array([points[index,0]+0.5, points[index,1]+0.5, points[index,2]+0.5])
    faces[index*6]= np.array([len(points)+8*index, len(points)+8*index+1,len(points)+8*index+2,len(points)+8*index+3])
    faces[index*6+1]= np.array([len(points)+8*index+4, len(points)+8*index+5,len(points)+8*index+6,len(points)+8*index+7])
    faces[index*6+2]= np.array([len(points)+8*index+2, len(points)+8*index+3,len(points)+8*index+6,len(points)+8*index+7])
    faces[index*6+3]= np.array([len(points)+8*index+0, len(points)+8*index+1,len(points)+8*index+4,len(points)+8*index+5])
    faces[index*6+4]= np.array([len(points)+8*index+0, len(points)+8*index+2,len(points)+8*index+4,len(points)+8*index+6])
    faces[index*6+5]= np.array([len(points)+8*index+1, len(points)+8*index+3,len(points)+8*index+5,len(points)+8*index+7])
return corners, faces

def writeocc(coordinates,save_path,filename):
  points = occ2points(coordinates)
  #print(points.shape)
  corners, faces = generate_faces(points)
  if points.shape[0] == 0:
      print('the predicted mesh has zero point!')
  else:
      points = np.concatenate((points,corners),axis=0)
      write_ply(points, faces, os.path.join(save_path,filename))

''' Just like this picture:visualization of voxel occupancy Thanks guys!

Answer 1

It looks like your code just need a small change to ensure that faces are stored preserving coherency in the orientation.

Replace the definition of faces in your generate_faces(points)function with this:

    base=len(points)+8*index
    faces[index*6]= np.array([base+2, base+3,base+1,base+0])
    faces[index*6+1]= np.array([base+4, base+5, base+7,base+6])
    faces[index*6+2]= np.array([base+3, base+2, base+6,base+7])
    faces[index*6+3]= np.array([base+0, base+1, base+5,base+4])
    faces[index*6+4]= np.array([base+2, base+0,base+4,base+6])
    faces[index*6+5]= np.array([base+1, base+3,base+7,base+5])
    
  return corners, faces

This is the result for a pair of voxels...