How to remove a polyline from an RGB image using numpy - python

I am not sure what would be the best way to remove a vertical polyline from an image. This line starts from the very top of an image to the bottom. The input is an image with three channels and array of y indices of the line. The output would be the same input image but without the polyline. For example, if the first image shape is (5, 10,3), then the final image shape should be (5, 9, 3) after removing the polyline.
Input example:
Image:
np.array([[[0, 20, 0, 0, 255, 0, 0, 0, 0, 0],
[0, 20, 0, 0, 255, 0, 0, 0, 0, 0],
[0, 56, 0, 0, 0, 255, 0, 0, 0, 0],
[0, 58, 0, 0, 0, 255, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 255, 0, 0, 0]],
[[0, 0, 0, 0, 255, 0, 0, 0, 30, 0],
[0, 0, 0, 0, 255, 0, 0, 0, 31, 0],
[0, 0, 0, 0, 0, 255, 0, 0, 32, 0],
[0, 0, 0, 0, 0, 255, 0, 0, 31, 0],
[0, 0, 0, 0, 0, 0, 255, 0, 0, 0]],
[[1, 0, 0, 0, 255, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 255, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 255, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 255, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 255, 0, 0, 0]]])
Note: the size here is (3, 5, 10) for illustration. The image shape would be (5, 10, 3). We wanna remove the white polyline
Polyline:
np.array([4, 4, 5, 5, 6])
Output Example
np.array([[[0, 20, 0, 0, 0, 0, 0, 0, 0],
[0, 20, 0, 0, 0, 0, 0, 0, 0],
[0, 56, 0, 0, 0, 0, 0, 0, 0],
[0, 58, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0]],
[[0, 0, 0, 0, 0, 0, 0, 30, 0],
[0, 0, 0, 0, 0, 0, 0, 31, 0],
[0, 0, 0, 0, 0, 0, 0, 32, 0],
[0, 0, 0, 0, 0, 0, 0, 31, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0]],
[[1, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0]]])
Note: the size now is (3, 5, 9)
My code
def remove_polyline(image, polyline):
h, w = image.shape[:2]
res_image = np.zeros((h, w - 1))
for i in range(h):
left_section = image[i, :polyline[i]]
right_section = image[i, polyline[i]+1:]
res_image[i, :] = np.dstack((left_section, right_section))
return res_image
I know this would work for grayscale images but not sure about 3 channel images


Note that your input has shape (3,5,10) instead of (5,10,3). This answer assumes the latter:
def remove_polyline(s, polyline):
mask = np.arange(s.shape[1]) != polyline[ :, None]
return s[mask,:].reshape(s[:,:-1].shape).copy()
r = remove_polyline(image, polyline)
plt.imshow(r)
Output:


You can try this.
def remove_polyline(image, polyline):
h, w, c = image.shape
res_image = np.zeros((h, w - 1, c))
for i in range(h):
left_section = image[i, :polyline[i], :]
right_section = image[i, polyline[i]+1:, :]
res_image[i, :, :] = np.vstack((left_section, right_section))
return res_image

Related

Have created a 3D array of chess boards, how can I add more boards to this?

neighboringStates = np.array([
[[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],
[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],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0]]
])
Wondering how I can add another 8x8 of zeros and ones and add it to this already existing 3D array. Thanks!

Use np.concatenate() to add a new array to the existing 3D array
Example Code:
import numpy as np
neighboringStates = np.array([
[[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],
[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],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0]]
])
new_array = np.array([
[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, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 1]
])
neighboringStates = np.concatenate((neighboringStates, np.array([new_array])))

Looping and counting python 2d arrays

I have array like this:
[
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0, 0]
]
I want to count value every 3 array, so the result i expected is:
[
[3, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 3, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 2, 1, 0, 0, 0, 0, 0]
]
I have no idea to loop it.
UPDATE..
this problem was solved. Thank you. I try Shijith's code this
if len(arr)%3==0:
print([[sum(y) for y in zip(arr[x],arr[x+1],arr[x+2])] for x in range(0, len(arr),3)])

Try:
result = []
for i in range(int(len(a)/3)):
result.append(np.sum(a[i*3:i*3+3], axis=0))
[array([3, 0, 0, 0, 0, 0, 0, 0, 0]),
array([0, 3, 0, 0, 0, 0, 0, 0, 0]),
array([0, 0, 2, 1, 0, 0, 0, 0, 0])]

you can use numpy.sum() along axis=0 , for every three rows in your array.
import numpy as np
if len(arr)%3==0:
print(np.array([np.sum(arr[x:x+3], axis = 0) for x in range(0, len(arr),3) ]))
[[3 0 0 0 0 0 0 0 0]
[0 3 0 0 0 0 0 0 0]
[0 0 2 1 0 0 0 0 0]]
or use simple list comprehension,
if len(arr)%3==0:
print([[sum(y) for y in zip(arr[x],arr[x+1],arr[x+2])] for x in range(0, len(arr),3)])
[[3, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 3, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 2, 1, 0, 0, 0, 0, 0]]

You can do this in pandas like this ( It splits the rows into 3, then takes the sum of each set of rows) :
import pandas as pd
import numpy as np
df=pd.DataFrame([
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0, 0]
])
df.groupby(np.arange(len(df))//3).sum().to_numpy().tolist()
output:
[[3, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 3, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 2, 1, 0, 0, 0, 0, 0]]
For a pure non import way:
combine=[]
for x in range(3):
combine.append(list(sum((a[x*3:x*3+3]))))
list(combine)
output:
[[3, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 3, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 2, 1, 0, 0, 0, 0, 0]]

result = [[sum(three) for three in
zip(arr[first], arr[first + 1], arr[first + 2])]
for first in range(0, len(array)-len(array)%3, 3)]
print(result)
Output
[[3, 0, 0, 0, 0, 0, 0, 0, 0], [0, 3, 0, 0, 0, 0, 0, 0, 0], [0, 0, 2, 1, 0, 0, 0, 0, 0]]

How to solve 4 x 4 array with 16 unknown with Python (Update)

I need to solve a 2 by 2 array with 4 unknown
A B
C D
I know all horizontal sum A+B=11, C+D=7
I know all vertical sum A+C=10, B+D=8
I know all diagonal sum A+D=15, B+C=3
I then use Python to solve for A,B,C,D
import numpy as np
A = [[1, 1, 1, 1],
[1, 0, 0, 1],
[1, 0, 1, 0],
[0, 0, 1, 1]]
a = [18, 15, 10, 7]
answera = np.linalg.solve(A, a)
print(answera)
And the answer is [9. 2. 1. 6.] which is correct
Now I need to solve 4 by 4 array with 16 unknown
A B C D
E F G H
I J K L
M N O P
I know horizontal sum A+B+C+D=10, E+F+G+H=26, I+J+K+L=42, M+N+O+P=58
I know vertical sum A+E+I+M=28, B+F+J+N=32, C+G+K+O=36, D+H+L+P=40
I know diagonal sum M=13, I+N=23, E+J+O=30, A+F+K+P=34, B+G+L=21, C+H=11, D=4
The other diagonal sum A=1, B+E=7, C+F+I=18, D+G+J+M=34, H+K+N=33, L+O=27, P=16
Which mean I know the value of the 4 corners.
I tried the following code but did not work
C = [[1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1],
[0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0],
[0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0],
[0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0],
[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]]
c = [10, 26, 42, 58, 7, 21, 39, 33, 27, 11, 23, 35, 30, 23, 32, 136]
answerc = np.linalg.solve(C, c)
print(answerc)
The correct answer should be [1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.] but I got error message
Traceback (most recent call last):
answerc = np.linalg.solve(C, c)
r = gufunc(a, b, signature=signature, extobj=extobj)
raise LinAlgError("Singular matrix")
numpy.linalg.LinAlgError: Singular matrix
Am I in the right direction? I will need to solve 5X5 with 25 unknown, 6X6 with 36 unknown and so on. Is there an easier way?
-----------------------------------------------------------------------------
Following Mr. Rory Daulton solution, I can solve the above 1 to 16 4X4 array without problem, but when I use it in another array with negative number, it doesn't give answer as expected;
The negative 4X4 array as follow
-20 -10 -5 0
-10 -20 -10 -5
-5 0 -10 -20
-10 -20 -10 -5
My python code as follow
import numpy as np
G = [[1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], # horizontal rows
[0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1],
[1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0], # vertical columns
[0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0],
[0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0],
[0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0], # forward diagonals
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0],
[1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1],
[0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 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], # back diagonals
[0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
]
g = [-35, -45, -35, -45, # horizontal rows
-45, -50, -35, -30, # vertical columns
-10, -25, -20, -55, -40, -10, 0, # forward diagonals
-20, -20, -30, -20, -35, -30, -5, # back diagonals
]
answerg = np.linalg.lstsq(G, g, rcond=None)
print(answerg[0])
The output is not exactly the original array
[-2.00000000e+01 -1.31250000e+01 -1.87500000e+00 8.88178420e-15
-6.87500000e+00 -2.00000000e+01 -1.00000000e+01 -8.12500000e+00
-8.12500000e+00 2.13162821e-14 -1.00000000e+01 -1.68750000e+01
-1.00000000e+01 -1.68750000e+01 -1.31250000e+01 -5.00000000e+00]
What should I try? Thank you in advance.

SHORT ANSWER: There are infinitely many solutions to your problem. So this takes a more complex analysis of the equations.
LONG ANSWER: You have multiple problems with your code.
First, you make it easy to make mistakes, since the lines of your matrix do not correspond to the data that you present. Worse, you have no comments to explain things. This mis-match will probably cause errors. You have 22 pieces of data in your sums, so use them. You tried to combine some of the sums and ignore others (the four corners) but you did not do it properly and you ended up with a singular matrix.
Next, you use linalg.solve. In your problem you have more data items (22) than unknowns (16), so solve is inappropriate. The numpy documentation for solve states
a must be square and of full-rank, i.e., all rows (or, equivalently,
columns) must be linearly independent; if either is not true, use
lstsq for the least-squares best “solution” of the system/equation.
The matrix resulting from your data is not square, therefore the rows are not linearly independent, so you should use lstsq rather than solve. The lstsq routine gives more information than you need for your problem, so just print the first item in the resulting list.
Combining those ideas and adding a few comments gives this code:
import numpy as np
C = [[1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], # horizontal rows
[0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1],
[1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0], # vertical columns
[0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0],
[0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0],
[0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0], # forward diagonals
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0],
[1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1],
[0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 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], # back diagonals
[0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
]
c = [10, 26, 42, 58, # horizontal rows
28, 32, 36, 40, # vertical columns
13, 23, 30, 34, 21, 11, 4, # forward diagonals
1, 7, 18, 34, 33, 27, 16, # back diagonals
]
answerc = np.linalg.lstsq(C, c, rcond=None)
print(answerc[0])
The printout is what you want:
[ 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.]
However, to be honest, there is no guarantee that this is an answer--just that it is a "closest" answer. Also, if it is an answer, there may be other answers. And, in fact, further analysis shows that there are other answers that satisfy all your conditions.
The sympy module can generate a row reduced echelon form of the matrix, which can be used to do more in-depth analysis of all the answers. However, the constants are then to be part of the matrix, rather than used as a separate array. Here is code for sympy to attempt to solve your problem:
import sympy
C = [[1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 10], # horizontal rows
[0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 26],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 42],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 58],
[1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 28], # vertical columns
[0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 32],
[0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 36],
[0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 40],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 13], # forward diagonals
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 23],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 30],
[1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 34],
[0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 21],
[0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 11],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4],
[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1], # back diagonals
[0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 7],
[0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 18],
[0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 34],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 33],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 27],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 16],
]
print(sympy.Matrix(C).rref())
The printout is
(Matrix([
[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, -1, 0, -13],
[0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 18],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 20],
[0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 6],
[0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 7],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, -1, 0, -7],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, -1, 0, -6],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 10],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 11],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 27],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 13],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 29],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 16],
[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, 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, 0, 0, 0, 0, 0, 0, 0]]), (0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15))
If you know how to analyze this you will see that there are infinitely many answers to your problem. If you set the number in the 15th cell to 15+x then the above matrix shows that the answer to all your restrictions is
1 2+x 3-x 4
5-x 6 7 8+x
9+x 10 11 12-x
13 14-x 15+x 16
The solve function of numpy only works if there is just one solution, so even if you had adjusted your matrix differently it would not have worked for you.
ANSWER TO YOUR UPDATE:
It seems that you missed the point of my answer. Your 4x4 problem has infinitely many answers, so there is no procedure that can choose the particular answer that you have in mind. The np.linalg.lstsq routine can find one of the answers to your problem but probably will not find your desired answer. You should consider it to be a coincidence that using that routine in your first problem gave your desired answer--that will probably not work in other problems.
It is a little hard to interpret the given answer to your new problem, since the scientific notation is hard to read. But all those matrix values are exact, and here they are as rational numbers in a particular format that should be obvious:
-20 -10-(3+1/8) - 5+(3+1/8) 0
-10+(3+1/8) -20 -10 - 5-(3+1/8)
- 5-(3+1/8) 0 -10 -20+(3+1/8)
-10 -20+(3+1/8) -10-(3+1/8) - 5
You see that the numpy's answer is the one that you expected, with the value 3+1/8 added to or subtracted from half the array values. This makes x=3+1/8 in the general answer that I gave you for your first problem.
This is as good as you can expect. Numpy gave you a correct answer--it has no idea how to choose the answer that you had in your head from the infinitely many correct answers to your problem. The only way to get just one answer is to change your problem--perhaps state the value in the first row and second column, or the sum of the first and third values in any one of the rows, or something similar.

Create boolean mask on TensorFlow

Suppose I have a list
x = [0, 1, 3, 5]
And I want to get a tensor with dimensions
s = (10, 7)
Such that the first column of the rows with indexes defined in x are 1, and 0 otherwise.
For this particular example, I want to obtain the tensor containing:
T = [[1, 0, 0, 0, 0, 0, 0],
[1, 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],
[1, 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]]
Using numpy, this would be the equivalent:
t = np.zeros(s)
t[x, 0] = 1
I found this related answer, but it doesn't really solve my problem.

Try this:
import tensorflow as tf
indices = tf.constant([[0, 1],[3, 5]], dtype=tf.int64)
values = tf.constant([1, 1])
s = (10, 7)
st = tf.SparseTensor(indices, values, s)
st_ordered = tf.sparse_reorder(st)
result = tf.sparse_tensor_to_dense(st_ordered)
sess = tf.Session()
sess.run(result)
Here is the output:
array([[0, 1, 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, 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]], dtype=int32)
I slightly modified your indexes so you can see the x,y format of the indices
To obtain what you originally asked, set:
indices = tf.constant([[0, 0], [1, 0],[3, 0], [5, 0]], dtype=tf.int64)
Output:
array([[1, 0, 0, 0, 0, 0, 0],
[1, 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],
[1, 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]], dtype=int32)

What is the difference between [[0 for _ in range(10)] for _ in range(10)] and [0 for _ in range(10)] * 10? [duplicate]

This question already has answers here:
Nested List Indices [duplicate]
(2 answers)
Closed 9 years ago.
>>> CM = [[0 for _ in range(10)]] * 10
>>> CM
[[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, 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]]
>>> CM[0][0] = CM[0][0] + 1
>>> CM
[[1, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0, 0, 0, 0, 0]]
I was trying to create a confusion matrix. It basically contains count of the (i, j) pairs.
I first created a list of lists, and then incremented the appropriate variable. However, it didn't work as expected. CM[i][0] got incremented for all values of i.
I found a work around.
>>> CM = [[0 for _ in range(10)] for _ in range(10)]
>>> CM
[[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, 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]]
>>> CM[0][0] = CM[0][0] + 1
>>> CM
[[1, 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, 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]]
But I would be grateful if someone could explain why the first method failed.

>>> CM = [[0 for _ in range(10)]] * 10
Is copying a reference to the same object, ten times. It is equivalent to this:
>>> x = [0 for _ in range(10)]
>>> CM = [x, x, x, x, x, x, x, x, x, x]
So manipulating one element causes side effects. Your workaround is elegant and correct.
Note:
This occurs since the elements of the lists are lists (which are mutable). If they were strings for example, which are immutable, it wouldn't be an issue if the same string was referenced in different lists, since they can't be manipulated. Python doesn't like to waste memory (unless explicitly told to ie. deepcopy), so copying lists will simply copy their references.

Categories