TensorFlow to use their own training good CPKT model, test identification _ depth Learning

Go out and talk about how to use TensorFlow to generate your own picture training model CPKT. This section describes how to use a trained CPKT model for test recognition.

Direct Line Code:

############################################################################################ #!/usr/bin/ python2.7 #-*-Coding:utf-8-*-#Author: Zhaoqinghui #Date: 2016.5.10 #Function: Test identification using the CPKT model ########## ################################################################################ Import TensorFlow as TF import  NumPy as NP import sys import OS #import math import cv2 from scipy import ndimage from skimage.filter.thresholding Import Threshold_adaptive Import Saveandrestorechar ######################################################### result_ label_index= ' Result_label.txt ' test_img_path= "./testimage/test" classnum=36 #####################################
	#################### #------Centric Operation--------def Getbestshift (img): CY,CX = Ndimage.measurements.center_of_mass (IMG) Rows,cols = Img.shape Shiftx = Np.round (COLS/2.0-CX). Astype (int) shifty = Np.round (rows/2.0-cy). Astype (int) return SHIF TX, Shifty def shift (Img,shiftx,shifty): rows, cols = Img.shape M = np.fLoat32 ([[[1,0,shiftx],[0,1,shifty]]]) shifted = Cv2.warpaffine (Img,m, (cols,rows)) return shifted #------ Read the corresponding number on the label------Def read_result_label_list (): Result_label_dir = [] Result_label = [] reader = open (Result_labe L_index) While 1:line = Reader.readline () tmp = Line.split ("") If not line:break Result_label_dir.append (tm P[1][0:-1]) for I in range (int (classnum)): Result_label.append (str (result_label_dir[i)) return Result_label # ------The recognition part of the picture------def processimageproposal (imageID): Print Test_img_path+str (imageID) + '. png ' Gray = cv2.imread (t Est_img_path+str (imageID) + '. png ', Cv2. Imread_grayscale) result_label=read_result_label_list () shiftx, shifty = Getbestshift (gray) Gray = shift (GRAY,SHIFTX , shifty) Flatten = Gray.flatten ()/255.0 pred = SaveandRestoreChar.sess.run (saveandrestorechar.predict,feed_dict={ saveandrestorechar.x:[flatten],saveandrestorechar.keep_prob:1.0}) print "Prediction:", int (pred[1]) Cv2.imshow (St R (imageID) +". png", Gray) Cv2.waitkey (0) def main (Argvs): imageID = argvs[1] #此处为参数设置 #imageid =12 print "image_id:", imageID PR
 Ocessimageproposal (imageID) if __name__ = = ' __main__ ': Main (SYS.ARGV)

Where result_label.txt is the correct result label; Test_img_path is the path to the test picture, using Result_label.txt to get the correct rate. Where the Saveandrestorechar file is:

############################################################################################ #!/usr/bin/ python2.7 #-*-Coding:utf-8-*-#Author: Zhaoqinghui #Date: 2016.5.10 #Function: Restore the model from CH Eckpoint ########################################################################################## Import TensorFlow as TF import numpy as NP import math classnum=36 x = Tf.placeholder (tf.float32,[none,28*28]) Y_ = Tf.placeholde R (Tf.float32,[none,classnum]) def weight_variable (shape): init = tf.truncated_normal (Shape,stddev = 0.1) return TF. Variable (INIT) def bias_variable (shape): init = tf.constant (0.1, shape = shape) return TF. Variable (INIT) # # Declare convolution operations and pool operations # The convolution operation declared here is a vanilla version with a step length of 1,padding of 0 # # Pool operation is a 2x2 max Pool def conv2d (x,w): # Strides: [Batch, In_height, In_width, In_channels] return tf.nn.conv2d (x,w,strides = [1,1,1,1],padding = ' SAME ') def maxpool2d (x) : Return Tf.nn.max_pool (x,ksize = [1,2,2,1], strides = [1,2,2,1],padding = ' SAME ') ## model Build Process # The first layer is [one convolution to a max pooling], the patch_size of the convolution layer is 5x5, the number of channels entered is 1 (because it is a grayscale map), the output is 32 feature maps # [5,5,1,32]: Patch_si Ze is 5x5, the number of input channels is 1, the number of output channels is 32 (here 32 is based on network definition, not calculated) X_image = Tf.reshape (x,[-1,28,28,1)) #把变成需要的格式 w_conv1 = Weight_ Variable ([5,5,1,32]) b_conv1 = Bias_variable ([#做相应的操作]), conv, relu, Maxpool h_conv1 = Tf.nn.relu (conv2d (x_image,w_ CONV1) +b_conv1) h_pool1 = maxpool2d (H_CONV1) # second tier [one convolution plus one maxpool] w_conv2 = weight_variable ([5,5,32,64]) B_conv2 = Bias_ Variable ([]) H_conv2 = Tf.nn.relu (conv2d (h_pool1,w_conv2) +b_conv2) h_pool2 = maxpool2d (h_conv2) # Full-join layer, total 1024 neurons, At this time the picture has two 2x2 maxpool, each step is 2, at this time the picture has become 7x7 w_fc1 = weight_variable ([7*7*64,1024]) B_FC1 = bias_variable ([1024]) H_ Pool2_flat = Tf.reshape (h_pool2,[-1,7*7*64]) H_fc1 = Tf.nn.relu (Tf.matmul (H_POOL2_FLAT,W_FC1) +b_fc1) # Add dropout in training,
Remember to close the test when ... # Keep_prob represents the possibility of a reserved parameter, when equal to 1.0 means not to dropout Keep_prob = Tf.placeholder ("float") #要输入的值 #keep_prob = 1 H_fc1_drop = tf.nn. Dropout (H_fc1,keep_prob) #添加 SoftmaX-Layer W_FC2 = Weight_variable ([1024,classnum]) B_FC2 = Bias_variable ([classnum]) Y_conv = Tf.nn.softmax (Tf.matmul (h_fc1_dr OP,W_FC2) # +B_FC2) # Loss cross_entropy =-tf.reduce_sum (Y_*tf.log (y_conv)) Train_step = Tf.train.AdamOptimizer (1e-4). MI Nimize (cross_entropy) correct_prediction = Tf.equal (Tf.argmax (y_conv,1), Tf.argmax (y_,1)) accuracy = Tf.reduce_mean ( Tf.cast (correct_prediction, "float")) predict = [Tf.reduce_max (Y_conv), Tf.argmax (y_conv,1) [0]] Saver = tf.train.Saver () Checkpoint_dir = "./tmp/train_model.cpkt" Sess = tf.

 InteractiveSession () Saver.restore (Sess,checkpoint_dir)

The simple tensorflow should be so successful with your own picture set. Have a good time ing, go on.