我有一个2D数组,a
,包含一组100 x,y,z坐标:
[[ 0.81 0.23 0.52]
[ 0.63 0.45 0.13]
...
[ 0.51 0.41 0.65]]
我想创建一个3D二进制图像b
,每个x,y,z维度有101个像素,坐标范围在0.00到1.00之间。
由a
定义的位置处的像素值应为1,所有其他像素的值应为0。
我可以使用b = np.zeros((101,101,101))
创建一个正确形状的零数组,但是如何使用a
为其创建坐标和切片?
答案 0 :(得分:4)
你可以这样做 -
# Get the XYZ indices
idx = np.round(100 * a).astype(int)
# Initialize o/p array
b = np.zeros((101,101,101))
# Assign into o/p array based on linear index equivalents from indices array
np.put(b,np.ravel_multi_index(idx.T,b.shape),1)
分配部分的运行时 -
让我们使用更大的网格进行计时。
In [82]: # Setup input and get indices array
...: a = np.random.randint(0,401,(100000,3))/400.0
...: idx = np.round(400 * a).astype(int)
...:
In [83]: b = np.zeros((401,401,401))
In [84]: %timeit b[list(idx.T)] = 1 #@Praveen soln
The slowest run took 42.16 times longer than the fastest. This could mean that an intermediate result is being cached.
1 loop, best of 3: 6.28 ms per loop
In [85]: b = np.zeros((401,401,401))
In [86]: %timeit np.put(b,np.ravel_multi_index(idx.T,b.shape),1) # From this post
The slowest run took 45.34 times longer than the fastest. This could mean that an intermediate result is being cached.
1 loop, best of 3: 5.71 ms per loop
In [87]: b = np.zeros((401,401,401))
In [88]: %timeit b[idx[:,0],idx[:,1],idx[:,2]] = 1 #Subscripted indexing
The slowest run took 40.48 times longer than the fastest. This could mean that an intermediate result is being cached.
1 loop, best of 3: 6.38 ms per loop
答案 1 :(得分:4)
首先,通过安全地将浮子四舍五入到整数开始。在上下文中,请参阅this问题。
a_indices = np.rint(a * 100).astype(int)
接下来,将b
中的索引指定为1.但是请注意使用普通list
而不是数组,否则您将触发index arrays的使用。似乎这种方法的性能与替代品的性能相当(感谢@Divakar!: - )
b[list(a_indices.T)] = 1
我创建了一个小尺寸10而不是100的小示例,以及2个尺寸而不是3个,以说明:
>>> a = np.array([[0.8, 0.2], [0.6, 0.4], [0.5, 0.6]])
>>> a_indices = np.rint(a * 10).astype(int)
>>> b = np.zeros((10, 10))
>>> b[list(a_indices.T)] = 1
>>> print(b)
[[ 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
[ 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
[ 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
[ 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
[ 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
[ 0. 0. 0. 0. 0. 0. 1. 0. 0. 0.]
[ 0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]
[ 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
[ 0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]
[ 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]]