The 6.3VGGnet study of "TensorFlow combat"

This is my rewrite of the code, can run, but over-fitting phenomenon is serious, do not know how to modify the better

#-*-Coding:utf-8-*-"" "Created on Wed Dec 14:45:35 2017@author:administrator" "" #coding: utf-8# Copyright the Te Nsorflow Authors. All rights reserved.## Licensed under the Apache License, Version 2.0 (the "License");  n compliance with the license.# obtain a copy of the License at## http://www.apache.org/licenses/LICENSE-2.0## Unless required by applicable law or agreed to in writing, software# distributed under the License was distributed on an " As is ' basis,# without warranties or CONDITIONS of any KIND, either express or implied.# see the License for the specific Language governing permissions and# limitations under the license.# ================================================== ============================import TensorFlow as Tfimport numpy as Npdata_name = ' yaleb_32x32.mat ' sele_num = 10import mat Lab.engineeng = Matlab.engine.start_matlab () t = Eng.data_imread_mse (data_name,sele_num) eng.quit () #t = Np.array (t) Train_ma = Np.arraY (T[0]). Astype (np.float32) Train_lab = Np.array (T[1]). Astype (np.int8) Test_ma = Np.array (T[2]). Astype (Np.float32) Test_lab = Np.array (T[3]). Astype (np.int8) Num_fea = Train_ma.shape[1]num_class = Train_lab.shape[1]image_row = 32imag E_column = 32def conv_op (input_op, name, KH, KW, n_out, DH, DW, p): n_in = Input_op.get_shape () [ -1].value with Tf.na Me_scope (name) as Scope:kernel = Tf.get_variable (scope+ "W", Shape=[kh, KW, N_in, n _out], Dtype=tf.float32, Initializer=tf.contrib.layers.xa Vier_initializer_conv2d ()) conv = tf.nn.conv2d (Input_op, Kernel, (1, DH, DW, 1), padding= ' same ') bias_init_v Al = Tf.constant (0.0, shape=[n_out], dtype=tf.float32) biases = tf. Variable (Bias_init_val, trainable=true, name= ' b ') z = tf.nn.bias_add (conv, biases) activation = Tf.nn.relu (z , name=scope) p + = [kernel, biases] return activation# full join layer function def FC_op (input_op, Name, N_out, p): n_in = Input_op.get_shape () [ -1].value with Tf.name_scope (name) as Scope:kerne L = tf.get_variable (scope+ "W", Shape=[n_in, N_out], dtype= Tf.float32, Initializer=tf.contrib.layers.xavier_initializer ()) biases = tf. Variable (Tf.constant (0.1, shape=[n_out], dtype=tf.float32), name= ' b ') activation = Tf.nn.relu_layer (Input_op, Kerne    L, biases, name=scope) p + = [kernel, biases] return activationdef mpool_op (input_op, name, kh, KW, DH, DW):  Return Tf.nn.max_pool (Input_op, ksize=[1, KH, KW, 1], strides=[1, DH, DW, 1], padding= ' same ', name=name) # assume Input_op shape is 224x22 4x3sess = tf. InteractiveSession () #----------define input and output---------------#x = Tf.placeholder (Tf.float32, [None, Num_fea]) Y_ = Tf.placeho Lder (Tf.float32, [None, Num_class]) X_image = Tf.reshape (x, [ -1,image_row,image_column,1]) Keep_prob = Tf.placeholder (tf.float32) # block 1 --Outputs 112x112x64p = []conv1_1 = Conv_op (X_image, name= "Conv1_1", Kh=3, Kw=3, n_out=64, Dh=1, Dw=1, p=p) conv1_2 = con V_op (Conv1_1, name= "Conv1_2", Kh=3, Kw=3, n_out=64, Dh=1, Dw=1, p=p) pool1 = Mpool_op (conv1_2, name= "Pool1", kh=2, K w=2, dw=2, dh=2) # block 2--Outputs 56x56x128conv2_1 = Conv_op (pool1, name= "Conv2_1", Kh=3, Kw=3, n_out=128, Dh=1, dw= 1, p=p) conv2_2 = Conv_op (Conv2_1, name= "Conv2_2", Kh=3, Kw=3, n_out=128, Dh=1, Dw=1, p=p) pool2 = Mpool_op (conv2_2, NA Me= "Pool2", kh=2, kw=2, dh=2, dw=2) # # block 3--Outputs 28x28x256conv3_1 = Conv_op (pool2, name= "Conv3_1", Kh=3, kw=  3, n_out=256, Dh=1, Dw=1, p=p) conv3_2 = Conv_op (Conv3_1, name= "Conv3_2", Kh=3, Kw=3, n_out=256, Dh=1, Dw=1, p=p) Conv3_3 =   Conv_op (Conv3_2, name= "Conv3_3", Kh=3, Kw=3, n_out=256, Dh=1, Dw=1, p=p) pool3 = Mpool_op (Conv3_3, name= "Pool3", kh=2, kw=2, dh=2, dw=2# Block 4--Outputs 14x14x512conv4_1 = Conv_op (pool3, name= "Conv4_1", Kh=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) conv4_2 = Conv_op (Conv4_1, name= "Conv4_2", Kh=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) Conv4_3 = Conv_op (Conv4_2, name= "Conv4_3", kh= 3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) Pool4 = Mpool_op (Conv4_3, name= "Pool4", kh=2, kw=2, dh=2, dw=2) # block 5--OUTP  UTS 7x7x512conv5_1 = Conv_op (Pool4, name= "Conv5_1", Kh=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) Conv5_2 = Conv_op (Conv5_1, Name= "Conv5_2", Kh=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) Conv5_3 = Conv_op (Conv5_2, name= "Conv5_3", Kh=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) pool5 = Mpool_op (Conv5_3, name= "Pool5", kh=2, kw=2, dw=2, dh=2) # flattenshp = Pool5.get_shape () Flatt  Ened_shape = Shp[1].value * Shp[2].value * shp[3].valueresh1 = Tf.reshape (POOL5, [-1, Flattened_shape], name= "Resh1") # Fully connectedfc6 = Fc_op (Resh1, name= "Fc6", n_out=4096, p=p) Fc6_drop = Tf.nn.dropout (Fc6, Keep_prob, name= "Fc6_drop") F C7 = Fc_op (Fc6_drop, name= "Fc7", n_out=4096, p=p) Fc7_drop = Tf.nn.dropout (Fc7, Keep_prob, name= "Fc7_drop") Fc8 = Fc_op (Fc7_drop, name= "Fc8", n_out =num_class, p=p) predictions = Tf.nn.softmax (FC8) cross_entropy = Tf.reduce_mean (-tf.reduce_sum (Y_ * Tf.log (predictions ), reduction_indices=[1]) Train_step = Tf.train.AdamOptimizer (1e-4). Minimize (cross_entropy) correct_prediction = Tf.equal (Tf.argmax (predictions,1), Tf.argmax (y_,1)) accuracy = Tf.reduce_mean (Tf.cast (Correct_prediction,        Tf.float32)) Tf.global_variables_initializer (). Run () for I in range (£): Train_accuracy = Accuracy.eval (feed_dict={ X:train_ma, Y_: Train_lab, keep_prob:1.0}) Print ("Step%d, training accuracy%g"% (I, train_accuracy)) Train_ste P.run (Feed_dict={x:train_ma, Y_: Train_lab, keep_prob:0.8}) print ("Test accuracy%g"%accuracy.eval (feed_dict={x:test _ma, Y_: Test_lab, keep_prob:1.0}))

Another, more convenient way to rewrite

#-*-Coding:utf-8-*-"" "Created on Wed Dec 15:40:44 2017@author:administrator" "" #-*-Coding:utf-8-*-"" "Created on Wed Dec 14:45:35 2017@author:administrator "" "#coding: utf-8# Copyright the TensorFlow Authors. All rights reserved.## Licensed under the Apache License, Version 2.0 (the "License");  n compliance with the license.# obtain a copy of the License at## http://www.apache.org/licenses/LICENSE-2.0## Unless required by applicable law or agreed to in writing, software# distributed under the License was distributed on an " As is ' basis,# without warranties or CONDITIONS of any KIND, either express or implied.# see the License for the specific Language governing permissions and# limitations under the license.# ================================================== ============================import TensorFlow as Tfimport numpy as Npdata_name = ' yaleb_32x32.mat ' sele_num = 10import mat Lab.engineeng = Matlab.engine.start_matlab () t= Eng.data_imread_mse (Data_name,sele_num) eng.quit () #t = Np.array (t) train_ma = Np.array (T[0]). Astype (Np.float32) Train_lab = Np.array (T[1]). Astype (np.int8) Test_ma = Np.array (T[2]). Astype (np.float32) Test_lab = Np.array (T[3]). Astype (np.int8) Num_fea = Train_ma.shape[1]num_class = Train_lab.shape[1]image_row = 32image_column = 32def Conv_op (in        PUT_OP, name, KH, KW, n_out, DH, DW, p): n_in = Input_op.get_shape () [ -1].value with Tf.name_scope (name) as scope:                                 Kernel = tf.get_variable (scope+ "W", Shape=[kh, KW, N_in, N_out],        Dtype=tf.float32, initializer=tf.contrib.layers.xavier_initializer_conv2d ()) Conv = tf.nn.conv2d (Input_op, Kernel, (1, DH, DW, 1), padding= ' same ') Bias_init_val = tf.constant (0.0, shape =[n_out], dtype=tf.float32) biases = tf. Variable (Bias_init_val, trainable=true, name= ' b ') z = tf.nn.bias_add (conv, biases) activation = TF.NN.RElu (z, name=scope) p + = [kernel, biases] return activation# full join layer function def fc_op (input_op, Name, N_out, p): N_                                 in = Input_op.get_shape () [ -1].value with Tf.name_scope (name) as Scope:kernel = Tf.get_variable (scope+ "W",                                  Shape=[n_in, N_out], Dtype=tf.float32, Initializer=tf.contrib.layers.xavier_initializer ()) biases = tf. Variable (Tf.constant (0.1, shape=[n_out], dtype=tf.float32), name= ' b ') activation = Tf.nn.relu_layer (Input_op, Kerne    L, biases, name=scope) p + = [kernel, biases] return activationdef mpool_op (input_op, name, kh, KW, DH, DW):  Return Tf.nn.max_pool (Input_op, ksize=[1, KH, KW, 1], strides=[1, DH, DW, 1], padding= ' same ', name=name) # assume Input_op shape is 224x22 4X3 # block 1--outputs 112x112x64def inference_op (Input_op, KEep_prob): P = [] # assume Input_op shape is 224x224x3 # block 1--outputs 112x112x64 conv1_1 = conv_op (input _op, Name= "Conv1_1", Kh=3, Kw=3, n_out=64, Dh=1, Dw=1, p=p) conv1_2 = Conv_op (conv1_1, name= "Conv1_2", Kh=3, Kw=3, N_o Ut=64, Dh=1, Dw=1, p=p) pool1 = Mpool_op (conv1_2, name= "Pool1", kh=2, kw=2, dw=2, dh=2) # block 2--outputs 56x 56x128 conv2_1 = Conv_op (pool1, name= "Conv2_1", Kh=3, Kw=3, n_out=128, Dh=1, Dw=1, p=p) conv2_2 = Conv_op (conv2_1 , name= "Conv2_2", Kh=3, Kw=3, n_out=128, Dh=1, Dw=1, p=p) pool2 = Mpool_op (conv2_2, name= "Pool2", kh=2, kw=2, dh=2 , dw=2) # # block 3--outputs 28x28x256 conv3_1 = Conv_op (pool2, name= "Conv3_1", Kh=3, Kw=3, n_out=256, Dh=1, DW =1, p=p) conv3_2 = Conv_op (Conv3_1, name= "Conv3_2", Kh=3, Kw=3, n_out=256, Dh=1, Dw=1, p=p) conv3_3 = Conv_op (conv3 _2, Name= "Conv3_3", Kh=3, Kw=3, n_out=256, Dh=1, Dw=1, p=p) pool3 = Mpool_op (Conv3_3, name= "Pool3", kh=2, kw=2 , dh=2, dw=2) # block 4--outputs 14x14x512 conv4_1 = Conv_op (pool3, name= "Conv4_1", Kh=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) conv4_2  = Conv_op (Conv4_1, name= "Conv4_2", Kh=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) Conv4_3 = Conv_op (Conv4_2, name= "Conv4_3",  Kh=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) Pool4 = Mpool_op (Conv4_3, name= "Pool4", kh=2, kw=2, dh=2, dw=2) # block 5--Outputs 7x7x512 Conv5_1 = Conv_op (Pool4, name= "Conv5_1", Kh=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) Conv5_2 = Conv_op (Conv5_1, name= "Conv5_2", Kh=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) Conv5_3 = Conv_op (Conv5_2, name= "Conv5_3", K     H=3, Kw=3, n_out=512, Dh=1, Dw=1, p=p) pool5 = Mpool_op (Conv5_3, name= "Pool5", kh=2, kw=2, dw=2, dh=2) # Flatten SHP = Pool5.get_shape () flattened_shape = Shp[1].value * Shp[2].value * Shp[3].value resh1 = Tf.reshape (POOL5, [ -1, Flattened_shape], name= "RESH1") # fully connected Fc6 = Fc_op (Resh1, name= "Fc6", n_out=4096, p=p) Fc6_drop = Tf.nn.dropout (Fc6, Keep_Prob, name= "Fc6_drop") Fc7 = Fc_op (Fc6_drop, name= "Fc7", n_out=4096, p=p) Fc7_drop = Tf.nn.dropout (fc7, Keep_prob, n Ame= "Fc7_drop") Fc8 = Fc_op (Fc7_drop, name= "Fc8", N_out=num_class, p=p) predictions = Tf.nn.softmax (FC8) return p Redictions, Fc8, p#----------define input and output---------------#sess = tf. InteractiveSession () x = Tf.placeholder (Tf.float32, [None, Num_fea]) Y_ = Tf.placeholder (Tf.float32, [None, Num_Class]) x _image = Tf.reshape (x, [ -1,image_row,image_column,1]) Keep_prob = Tf.placeholder (tf.float32) predictions, FC8, p = Inference_op (X_image, keep_prob) cross_entropy = Tf.reduce_mean (-tf.reduce_sum (y_) tf.log (predictions), reduction_ INDICES=[1]) Train_step = Tf.train.AdamOptimizer (1e-4). Minimize (cross_entropy) correct_prediction = Tf.equal ( Tf.argmax (predictions,1), Tf.argmax (y_,1)) accuracy = Tf.reduce_mean (Tf.cast (correct_prediction, Tf.float32)) Tf.global_variables_initializer (). Run () for I in range: Train_accuracy = accuracy.eval (feed_dict={x:train_m A, Y_: Train_lab, keep_prob:1.0}) Print ("Step%d, training accuracy%g"% (I, train_accuracy)) Train_step.run (feed_dict={x: Train_ma, Y_: Train_lab, keep_prob:0.8}) print ("Test accuracy%g"%accuracy.eval (feed_dict={x:test_ma, Y_: Test_Lab, keep_prob:1.0}))

　　

　　

The 6.3VGGnet study of "TensorFlow combat"