基于python实现FPS（最远距离采样）
基于python实现FPS（最远距离采样）
FPS的原理：
目前深度学习第一步encoder主要采用的是FPS。（PointNET++）
且FPS确实能在点云表面生成一组分布均匀的点。
import numpy as np
from plyfile import PlyData
import time
import matplotlib.pyplot as plt
def read_ply_points(ply_path):ply = PlyData.read(ply_path)data = ply.elements[0].datapoints = np.stack([data['x'], data['y'], data['z']], axis=1)return points
class FPS:def __init__(self, points):self.points = np.unique(points, axis=0)def get_min_distance(self, a, b):distance = []for i in range(a.shape[0]):dis = np.sum(np.square(a[i] - b), axis=-1)distance.append(dis)distance = np.stack(distance, axis=-1)distance = np.min(distance, axis=-1)return np.argmax(distance)@staticmethoddef get_model_corners(model):min_x, max_x = np.min(model[:, 0]), np.max(model[:, 0])min_y, max_y = np.min(model[:, 1]), np.max(model[:, 1])min_z, max_z = np.min(model[:, 2]), np.max(model[:, 2])corners_3d = np.array([[min_x, min_y, min_z],[min_x, min_y, max_z],[min_x, max_y, min_z],[min_x, max_y, max_z],[max_x, min_y, min_z],[max_x, min_y, max_z],[max_x, max_y, min_z],[max_x, max_y, max_z],])return corners_3ddef compute_fps(self, K):# 计算中心点位corner_3d = self.get_model_corners(self.points)center_3d = (np.max(corner_3d, 0) + np.min(corner_3d, 0)) / 2A = np.array([center_3d])B = np.array(self.points)t = []# 寻找Ｋ个节点for i in range(K):print(i)max_id = self.get_min_distance(A, B)A = np.append(A, np.array([B[max_id]]), 0)B = np.delete(B, max_id, 0)t.append(max_id)file=open('fps_of.ply','w')file.writelines("ply\n")file.writelines("format ascii 1.0\n")file.writelines("element vertex "+str(K+1)+"\n")file.writelines("property float x\n")file.writelines("property float y\n")file.writelines("property float z\n")file.writelines("element face 0\n")file.writelines("property list uchar int vertex_indices\n")file.writelines("end_header\n")for i in A:file.writelines(str(float(i[0]))+"\t"+str(float(i[1]))+"\t"+str(float(i[2]))+"\n")
print("开始计时")
time_start = time.time()
f = FPS(read_ply_points("fanDisk.ply"))
f.compute_fps(1500)#采样点个数
time_end = time.time()
time_c= time_end - time_start #运行所花时间
print('time cost', time_c, 's')
时间复杂度大，更新一下：输入指定路径，可以把该路径下所有文件做FPS（读入txt）
import os
# 遍历文件夹及其子文件夹中的文件，并存储在一个列表中
# 输入文件夹路径、空文件列表[]
# 返回 文件列表Filelist,包含文件名（完整路径）
def get_filelist(dir, Filelist):newDir = dirif os.path.isfile(dir):Filelist.append(dir)# # 若只是要返回文件文，使用这个# Filelist.append(os.path.basename(dir))elif os.path.isdir(dir):for s in os.listdir(dir):newDir = os.path.join(dir, s)get_filelist(newDir, Filelist)return Filelistimport numpy as np
olderr = np.seterr(all='ignore')
def farthest_point_sample(point, npoint):N, D = point.shapexyz = point[:,:3]centroids = np.zeros((npoint,))distance = np.ones((N,)) * 1e10farthest = np.random.randint(0, N)for i in range(npoint):centroids[i] = farthestcentroid = xyz[farthest, :]dist = np.sum((xyz - centroid) ** 2, -1)mask = dist < distancedistance[mask] = dist[mask]farthest = np.argmax(distance, -1)point = point[centroids.astype(np.int32)]return pointdef run_fps(path):point=np.loadtxt(path)sample_point = 2048 # 输入想要采样的点的个数pcd_point = farthest_point_sample(point, sample_point) # FPS采样#保存路径savepath="E:"+path.split(":")[1]file = open(savepath, 'w')for i in pcd_point:file.write(str(i[0]) + " " + str(i[1]) + " " + str(i[2]) + " " + str(i[3]) + " " + str(i[4]) + " " + str(i[5]) + " " + str((i[6])) + "\n")file.close()if __name__ == '__main__':list = get_filelist('d:\\cvpr', [])for e in list:print(e)run_fps(e)
基于python实现FPS（最远距离采样）
基于python实现FPS（最远距离采样）
FPS的原理：
目前深度学习第一步encoder主要采用的是FPS。（PointNET++）
且FPS确实能在点云表面生成一组分布均匀的点。
import numpy as np
from plyfile import PlyData
import time
import matplotlib.pyplot as plt
def read_ply_points(ply_path):ply = PlyData.read(ply_path)data = ply.elements[0].datapoints = np.stack([data['x'], data['y'], data['z']], axis=1)return points
class FPS:def __init__(self, points):self.points = np.unique(points, axis=0)def get_min_distance(self, a, b):distance = []for i in range(a.shape[0]):dis = np.sum(np.square(a[i] - b), axis=-1)distance.append(dis)distance = np.stack(distance, axis=-1)distance = np.min(distance, axis=-1)return np.argmax(distance)@staticmethoddef get_model_corners(model):min_x, max_x = np.min(model[:, 0]), np.max(model[:, 0])min_y, max_y = np.min(model[:, 1]), np.max(model[:, 1])min_z, max_z = np.min(model[:, 2]), np.max(model[:, 2])corners_3d = np.array([[min_x, min_y, min_z],[min_x, min_y, max_z],[min_x, max_y, min_z],[min_x, max_y, max_z],[max_x, min_y, min_z],[max_x, min_y, max_z],[max_x, max_y, min_z],[max_x, max_y, max_z],])return corners_3ddef compute_fps(self, K):# 计算中心点位corner_3d = self.get_model_corners(self.points)center_3d = (np.max(corner_3d, 0) + np.min(corner_3d, 0)) / 2A = np.array([center_3d])B = np.array(self.points)t = []# 寻找Ｋ个节点for i in range(K):print(i)max_id = self.get_min_distance(A, B)A = np.append(A, np.array([B[max_id]]), 0)B = np.delete(B, max_id, 0)t.append(max_id)file=open('fps_of.ply','w')file.writelines("ply\n")file.writelines("format ascii 1.0\n")file.writelines("element vertex "+str(K+1)+"\n")file.writelines("property float x\n")file.writelines("property float y\n")file.writelines("property float z\n")file.writelines("element face 0\n")file.writelines("property list uchar int vertex_indices\n")file.writelines("end_header\n")for i in A:file.writelines(str(float(i[0]))+"\t"+str(float(i[1]))+"\t"+str(float(i[2]))+"\n")
print("开始计时")
time_start = time.time()
f = FPS(read_ply_points("fanDisk.ply"))
f.compute_fps(1500)#采样点个数
time_end = time.time()
time_c= time_end - time_start #运行所花时间
print('time cost', time_c, 's')
时间复杂度大，更新一下：输入指定路径，可以把该路径下所有文件做FPS（读入txt）
import os
# 遍历文件夹及其子文件夹中的文件，并存储在一个列表中
# 输入文件夹路径、空文件列表[]
# 返回 文件列表Filelist,包含文件名（完整路径）
def get_filelist(dir, Filelist):newDir = dirif os.path.isfile(dir):Filelist.append(dir)# # 若只是要返回文件文，使用这个# Filelist.append(os.path.basename(dir))elif os.path.isdir(dir):for s in os.listdir(dir):newDir = os.path.join(dir, s)get_filelist(newDir, Filelist)return Filelistimport numpy as np
olderr = np.seterr(all='ignore')
def farthest_point_sample(point, npoint):N, D = point.shapexyz = point[:,:3]centroids = np.zeros((npoint,))distance = np.ones((N,)) * 1e10farthest = np.random.randint(0, N)for i in range(npoint):centroids[i] = farthestcentroid = xyz[farthest, :]dist = np.sum((xyz - centroid) ** 2, -1)mask = dist < distancedistance[mask] = dist[mask]farthest = np.argmax(distance, -1)point = point[centroids.astype(np.int32)]return pointdef run_fps(path):point=np.loadtxt(path)sample_point = 2048 # 输入想要采样的点的个数pcd_point = farthest_point_sample(point, sample_point) # FPS采样#保存路径savepath="E:"+path.split(":")[1]file = open(savepath, 'w')for i in pcd_point:file.write(str(i[0]) + " " + str(i[1]) + " " + str(i[2]) + " " + str(i[3]) + " " + str(i[4]) + " " + str(i[5]) + " " + str((i[6])) + "\n")file.close()if __name__ == '__main__':list = get_filelist('d:\\cvpr', [])for e in list:print(e)run_fps(e)