Tensorflow实现进阶的卷积网络

此部分代码来自于《Tensorflow实战》5.3节

#前两行代码可以跳过，自行去https://github.com/tensorflow/models/tree/master/tutorials/image/cifar10中将cifar10.py和cifar10_input.py下载下来'''git clone https://github.com/tensorflow/models.gitcd models/tutorials/image/cifar10'''import cifar10,cifar10_inputimport tensorflow as tfimport numpy as npimport timemax_steps = 3000batch_size = 128data_dir = 'D:/Sublime Text 3/cifar10_data/cifar-10-batches-bin'#此处存放的路径我写的绝对路径  书中写的/tmp应该是IDLE的安装位置def variable_with_weight_loss(shape, stddev, wl):var = tf.Variable(tf.truncated_normal(shape, stddev = stddev))if wl is not None:weight_loss = tf.multiply(tf.nn.l2_loss(var), wl, name = 'weight_loss')tf.add_to_collection('losses', weight_loss)return var#把cifar10的数据解压到data_dir中，然后将下一行代码注释掉，取消运行#cifar10.maybe_download_and_extract()images_train, labels_train = cifar10_input.distorted_inputs(data_dir = data_dir, batch_size = batch_size)#这行代码在Tensorflow 0.12版中会报错【TypeError: strided_slice() missing 1 required positional argument: 'strides'】我将版本升至1.0后就没有报错了images_test, labels_test = cifar10_input.inputs(eval_data = True, data_dir = data_dir, batch_size = batch_size)image_holder = tf.placeholder(tf.float32, [batch_size, 24, 24, 3])label_holder = tf.placeholder(tf.int32, [batch_size])weight1 = variable_with_weight_loss(shape = [5,5,3,64], stddev = 5e-2, wl = 0.0)kernel1 = tf.nn.conv2d(image_holder, weight1, [1,1,1,1], padding = 'SAME')bias1 = tf.Variable(tf.constant(0.0, shape = [64]))conv1 = tf.nn.relu(tf.nn.bias_add(kernel1, bias1))pool1 = tf.nn.max_pool(conv1, ksize = [1,3,3,1], strides = [1,2,2,1], padding = 'SAME')norm1 = tf.nn.lrn(pool1, 4, bias = 1.0, alpha = 0.001/9.0, beta = 0.75)weight2 = variable_with_weight_loss(shape = [5,5,64,64], stddev = 5e-2, wl = 0.0)kernel2 = tf.nn.conv2d(norm1, weight2, [1,1,1,1], padding = 'SAME')bias2 = tf.Variable(tf.constant(0.1, shape = [64]))conv2 = tf.nn.relu(tf.nn.bias_add(kernel2, bias2))norm2 = tf.nn.lrn(conv2, 4, bias = 1.0, alpha = 0.001/9.0, beta = 0.75)pool2 = tf.nn.max_pool(norm2, ksize = [1,3,3,1], strides = [1,2,2,1], padding = 'SAME')reshape = tf.reshape(pool2, [batch_size, -1])#比如pool2总共有a×b×c个数，要reshape成batch_size行 × ???列，这里的-1就表示???列，tf.reshape会自行算出需要多少列a×b×c/batch_sizedim = reshape.get_shape()[1].valueweight3 = variable_with_weight_loss(shape = [dim, 384], stddev = 0.04, wl = 0.004)bias3 = tf.Variable(tf.constant(0.1, shape = [384]))local3 = tf.nn.relu(tf.matmul(reshape, weight3) + bias3)weight4 = variable_with_weight_loss(shape = [384,192], stddev = 0.04, wl = 0.004)bias4 = tf.Variable(tf.constant(0.1, shape = [192]))local4 = tf.nn.relu(tf.matmul(local3, weight4) + bias4)weight5 = variable_with_weight_loss(shape = [192,10], stddev = 1/192.0, wl = 0.0)bias5 = tf.Variable(tf.constant(0.0, shape = [10]))logits = tf.nn.relu(tf.matmul(local4, weight5) + bias5)def loss(logits, labels):labels = tf.cast(labels, tf.int64)cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits = logits, labels = labels, name = 'cross_entropy_per_example')cross_entropy_mean = tf.reduce_mean(cross_entropy, name = 'cross_entropy')tf.add_to_collection('losses', cross_entropy_mean)return tf.add_n(tf.get_collection('losses'), name = 'total_loss')loss = loss(logits, label_holder)train_op = tf.train.AdamOptimizer(1e-3).minimize(loss)top_k_op = tf.nn.in_top_k(logits, label_holder, 1)sess = tf.InteractiveSession()tf.global_variables_initializer().run()tf.train.start_queue_runners()for step in range(max_steps):start_time = time.time()image_batch, label_batch = sess.run([images_train, labels_train])_, loss_value = sess.run([train_op, loss], feed_dict = {image_holder: image_batch, label_holder: label_batch})duration = time.time() - start_timeif step %10 ==0:examples_per_sec = batch_size / durationsec_per_batch = float(duration)format_str = ('step %d,loss=%.2f (%.1f example/sec; %.3f sec/batch)')print(format_str % (step, loss_value, examples_per_sec, sec_per_batch))num_examples = 10000import mathnum_iter = int(math.ceil(num_examples / batch_size))true_count = 0total_sample_count = num_iter * batch_sizestep = 0while step < num_iter:image_batch, label_batch = sess.run([images_test,labels_test])predictions = sess.run([top_k_op], feed_dict = {image_holder: image_batch, label_holder: label_batch})true_count += np.sum(predictions)step += 1precision = true_count / total_sample_countprint('precision @ 1 = %.3f' % precision)

max_step=100,num_example=100的运行结果（CPU版）

Filling queue with 20000 CIFAR images before starting to train. This will take a few minutes.step 0,loss=4.67 (26.4 example/sec; 4.854 sec/batch)step 10,loss=3.81 (127.4 example/sec; 1.005 sec/batch)step 20,loss=3.20 (118.6 example/sec; 1.079 sec/batch)step 30,loss=2.86 (128.0 example/sec; 1.000 sec/batch)step 40,loss=2.59 (125.0 example/sec; 1.024 sec/batch)step 50,loss=2.48 (125.8 example/sec; 1.018 sec/batch)step 60,loss=2.34 (126.3 example/sec; 1.013 sec/batch)step 70,loss=2.28 (128.4 example/sec; 0.997 sec/batch)step 80,loss=2.30 (128.4 example/sec; 0.997 sec/batch)step 90,loss=2.29 (127.8 example/sec; 1.002 sec/batch)precision @ 1 = 0.297[Finished in 108.8s]