The import OS 
The import struct 
The import numpy as np 
The import cv2 

Def load_mnist (path, kind='train') : 
"" "the Load MNIST data from ` path ` "" "
Labels_path=OS. Path. Join (path, 
'% s - labels. Idx1 - ubyte' 
% kind) 
Images_path=OS. Path. Join (path, 
'% s - images. Idx3 does - ubyte' 
% kind) 
With the open (labels_path, 'rb') as lbpath: 
Magic, n=struct. Unpack (' & gt; II ', 
Lbpath. Read (8)) 
Labels=np. Fromfile (lbpath, 
Dtype=np. Uint8) 

With the open (images_path, 'rb') as imgpath: 
Magic, num, rows, cols=struct. Unpack (' & gt; IIII ', 
Imgpath. Read (16)) 
Images=np. Fromfile (imgpath, 
Dtype=np. Uint8). Reshape (len (labels), 784) 

The return of images, labels 

The image and labels=load_mnist (r 'C: \ Users \ \ Desktop \ Python Image_processing \ data_number', '" train ") 

CharNum=10 
"' n of the data set update label format for n X 10 '"' 
TrainingLabel=[] 
For I in range (len (labels) : 
LabelArraytoAppend=[] 
For j in range (charNum) : 
If labels [I]==j: 
LabelArraytoAppend. Append (1) 
The else: 
LabelArraytoAppend. Append (0) 
TrainingLabel. Append (labelArraytoAppend) 

"' training multi-layer neural network MLP '
Ann=cv2. Ml. ANN_MLP_create () 
Ann. SetActivationFunction (cv2. Ml. ANN_MLP_SIGMOID_SYM) 
Ann. SetTrainMethod (cv2. Ml. ANN_MLP_BACKPROP) 
Ann. SetBackpropMomentumScale (0.1) 
Ann. SetBackpropWeightScale (0.001) 
Ann. SetTermCriteria ((cv2 TERM_CRITERIA_COUNT | cv2. TERM_CRITERIA_EPS, 500, 0.05)) 

Ann. SetLayerSizes (np) array ([np. Array (image). Shape [1], 128, charNum], dtype=np, uint16)) 

Ann. "train" (np) array (image, dtype=np. Float32), cv2. Ml. ROW_SAMPLE, np, array (trainingLabel, dtype=np. Float32)) 
Ann. Save (r 'C: \ Users \ bj0lt5 \ \ Desktop \ Python Image_processing \ data_number \ characterRecognize XML') 

"' test neural network" '
TestImage, testLabels=load_mnist (r 'C: \ Users \ \ Desktop \ Python Image_processing \ data_number', 't10k') 
# result=np. Zeros ((testLabels shape [0], charNum), dtype=np. Float32) 
Retval, result=Ann predict (np) array (image, dtype=np. Float32)) 
Print (result) [: 2] 

CodePudding user response:

Why did you use python without tensorflow, data are more

CodePudding user response:

New haven't come into contact with the Tensor flow,, I go to learn, but so should also can realize, don't know where is there a problem

CodePudding user response:

Hello ~ I want to ask you that you solve this problem? Because I now also in the use of opencv to train the neural network, I also encountered this problem,