Convert string back to numpy array

Question

I'am working on a video chat in python. I'am using CV2 to capture the current image(this gives me a numpy array), then I send the numpy over to my server. Now I have a string at the server and i need to decode it back to a numpy array.

I'am using Python 3.7 and i dindn't come up with somethink yet.

    #client 
    #capture 500 frames
    while i < 500:
        i = i + 1
        # Capture frame-by-frame
        ret, frame = cap.read()
        #send data                           
        client_socket.send(bytes(str(ret) + "###" + str(frame), "utf-8"))


    #server
    #split ret and frame
    ret, frame = str(conn.recv(16000)).split("###")
    gray = cv2.cvtColor(frame.toNumpyArray #PseudoMethod  <-- Here
    ,cv2.COLOR_BGR2GRAY)

I only need a method to convert the string back to a numpy array. If i print it out,the string looks like this:

b'[[[128 255 255]\n [125 255 255]\n [107 255 255]\n ...\n [102 130 167]\n [102 128 172]\n [102 128 172]]\n\n [[128 255 255]\n [127 255 255]\n [108 255 255]\n ...\n [102 130 167]\n [102 128 172]\n [102 128 172]]\n\n [[111 255 255]\n [111 255 255]\n [109 255 255]\n ...\n [ 99 131 169]\n [ 99 131 169]\n [ 99 131 169]]\n\n ...\n\n [[ 27 64 95]\n [ 29 67 97]\n [ 24 66 98]\n ...\n [ 73 117 160]\n [ 70 119 161]\n [ 70 119 161]]\n\n [[ 18 71 81]\n [ 20 74 83]\n [ 30 67 93]\n ...\n [ 77 117 159]\n [ 74 118 163]\n [ 74 118 163]]\n\n [[ 14 68 77]\n [ 19 73 82]\n [ 30 67 93]\n ...\n [ 77 117 159]\n [ 74 118 163]\n [ 74 118 163]]]'

Sorry for my bad english, I'am a german student.

It seems like you're looking for ways to [serialize](https://markhneedham.com/blog/2018/04/07/python-serialize-deserialize-numpy-2d-arrays/) — rb612, Jan 02 '19 at 09:43
See the '...'? That's missing data. You can't recover the original from this. `str` is the wrong way to serialize an array. — hpaulj, Jan 02 '19 at 11:54
See a related question and its answers here https://stackoverflow.com/questions/38886641/convert-string-representation-of-array-to-numpy-array-in-python — Hackless, Sep 22 '22 at 04:12

score 2 · Accepted Answer · answered Jan 02 '19 at 11:10

The following pair of programs demonstrates one way to commmunicate a numpy ndarray object across a network socket. The client converts the array to a byte stream using the save method, writing the stream to a BytesIO object that is then sent across the socket to the server:

import numpy as np
import socket
from io import BytesIO

# Create an output socket connected to server
sout = socket.create_connection(('127.0.0.1', 6543))

# Create data and write binary stream out to socket

a = np.array([[1.1, 2.2, 3.3],
              [4.4, 5.5, 6.6],
              [7.7, 8.8, 9.9]])

b = BytesIO()
np.save(b, a)

sout.send(b.getvalue())
sout.close()

The server listens on the appropriate network address, receiving data until the ending socket closes. It then converts the received data into a BytesIO object, from which numpy's load function recovers the structure:

import numpy as np
import socket
from io import BytesIO

# Create socket listening on correct port
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.bind(('127.0.0.1', 6543))
s.listen()

# Accept a connection and receive data
sock, address = s.accept()
data = b''
while True:
    indt = sock.recv(1024)
    if not indt:
        break
    data += indt

# Take data and recast to np.ndarray
data = BytesIO(data)

b = np.load(data)
print(type(b), b, sep='\n')

The output from running the server is as follows:

<class 'numpy.ndarray'>
[[1.1 2.2 3.3]
 [4.4 5.5 6.6]
 [7.7 8.8 9.9]]

There are various ways in which this code could be optimised, but this should give you enough to get going.

score 1 · Answer 2 · answered Jan 02 '19 at 09:48

1

I think that the numpy.fromstring function does exactly what you want.

For example the np.fromstring('1 2', dtype=int, sep=' ') call returns this array: array([1, 2]).

Hope it helps.

answered Jan 02 '19 at 09:48

Petros

342
3
15

That's not what his string looks like. – hpaulj Jan 02 '19 at 12:28

Convert string back to numpy array

2 Answers2

Linked