卷积网络博大精深，不同的网络模型，跑出来的结果是不一样，在不知道使用什么网络的情况下跑自己的数据集时，我建议最好去参考基于cnn的手写数字识别网络构建，在其基础上进行改进，对于一般测试数据集有很大的帮助。
分享一个网络构架和一中训练方法：
# coding:utf-8 import os import tensorflow as tf os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' # cnn模型高度抽象特征 def cnn_face_discern_model(X_,Y_): weights = { "wc1":tf.Variable(tf.random_normal([3,3,1,64],stddev=0.1)), "wc2":tf.Variable(tf.random_normal([5,5,64,128],stddev=0.1)), "wd3":tf.Variable(tf.random_normal([7*7*128,1024],stddev=0.1)), "wd4": tf.Variable(tf.random_normal([1024, 12], stddev=0.1)) } biases = { "bc1":tf.Variable(tf.random_normal([64],stddev=0.1)), "bc2":tf.Variable(tf.random_normal([128],stddev=0.1)), "bd3": tf.Variable(tf.random_normal([1024],stddev=0.1)), "bd4": tf.Variable(tf.random_normal([12],stddev=0.1)) } x_input = tf.reshape(X_,shape=[-1,28,28,1]) # 第一层卷积层 _conv1 = tf.nn.conv2d(x_input,weights["wc1"],strides=[1,1,1,1],padding="SAME") _conv1_ = tf.nn.relu(tf.nn.bias_add(_conv1,biases["bc1"])) # 第一层池化层 _pool1 = tf.nn.max_pool(_conv1_,ksize=[1,2,2,1],strides=[1,2,2,1],padding="SAME") # 第一层失活层 _pool1_dropout = tf.nn.dropout(_pool1,0.7) # 第二层卷积层 _conv2 = tf.nn.conv2d(_pool1_dropout,weights["wc2"],strides=[1,1,1,1],padding="SAME") _conv2_ = tf.nn.relu(tf.nn.bias_add(_conv2,biases["bc2"])) # 第二层池化层 _pool2 = tf.nn.max_pool(_conv2_,ksize=[1,2,2,1],strides=[1,2,2,1],padding="SAME") # 第二层失活层 _pool2_dropout = tf.nn.dropout(_pool2,0.7) # 使用全连接层提取抽象特征 # 全连接层1 _densel = tf.reshape(_pool2_dropout,[-1,weights["wd3"].get_shape().as_list()[0]]) _y1 = tf.nn.relu(tf.add(tf.matmul(_densel,weights["wd3"]),biases["bd3"])) _y2 = tf.nn.dropout(_y1,0.7) # 全连接层2 out = tf.add(tf.matmul(_y2,weights["wd4"]),biases["bd4"]) # 损失函数 loss loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(labels=Y_, logits=out)) # 计算交叉熵 # 优化目标 optimizing optimizing = tf.train.AdamOptimizer(0.001).minimize(loss) # 使用adam优化器来以0.0001的学习率来进行微调 # 精确度 accuracy correct_prediction = tf.equal(tf.argmax(Y_, 1), tf.argmax(out, 1)) # 判断预测标签和实际标签是否匹配 accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float")) return { "loss":loss, "optimizing":optimizing, "accuracy":accuracy, "out":out }
批量训练方法：
# 开始准备训练cnn X = tf.placeholder(tf.float32,[None,28,28,1]) # 这个12属于人脸类别，一共有几个id Y = tf.placeholder(tf.float32, [None,12]) # 实例化模型 cnn_model = cnn_face_discern_model(X,Y) loss,optimizing,accuracy,out = cnn_model["loss"],cnn_model["optimizing"],cnn_model["accuracy"],cnn_model["out"] # 启动训练模型 bsize = 960/60 with tf.Session() as sess: # 实例所有参数 sess.run(tf.global_variables_initializer()) for epoch in range(100): for i in range(15): x_bsize,y_bsize = x_train[i*60:i*60+60,:,:,:],y_train[i*60:i*60+60,:] sess.run(optimizing,feed_dict={X:x_bsize,Y:y_bsize}) if (epoch+1)%10==0: los = sess.run(loss,feed_dict={X:x_test,Y:y_test}) acc = sess.run(accuracy,feed_dict={X:x_test,Y:y_test}) print("epoch:%s loss:%s accuracy:%s"%(epoch,los,acc)) score= sess.run(accuracy,feed_dict={X:x_test,Y:y_test}) y_pred = sess.run(out,feed_dict={X:x_test}) # 这个是类别，测试集预测出来的类别。 y_pred = np.argmax(y_pred,axis=1) print("最后的精确度为：%s"%score)