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.

• Nitin Pawar: Can you provide an example of the string you're getting?
• Lukas Neumann: I have posted an example
• rb612: It seems like you're looking for ways to serialize
• hpaulj: See the '...'? That's missing data. You can't recover the original from this. str is the wrong way to serialize an array.
• Hackless: See a related question and its answers here stackoverflow.com/questions/38886641/…

Answer 1

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.

Answer 2

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.

• hpaulj: That's not what his string looks like.