在Tensorflow卷積神經(jīng)網(wǎng)絡(luò)實(shí)例這篇博客中,我們實(shí)現(xiàn)了一個(gè)簡(jiǎn)單的卷積神經(jīng)網(wǎng)絡(luò),沒(méi)有復(fù)雜的Trick。接下來(lái),我們將使用CIFAR-10數(shù)據(jù)集進(jìn)行訓(xùn)練。
CIFAR-10是一個(gè)經(jīng)典的數(shù)據(jù)集,包含60000張32*32的彩色圖像,其中訓(xùn)練集50000張,測(cè)試集10000張。CIFAR-10如同其名字,一共標(biāo)注為10類,每一類圖片6000張。
本文實(shí)現(xiàn)了進(jìn)階的卷積神經(jīng)網(wǎng)絡(luò)來(lái)解決CIFAR-10分類問(wèn)題,我們使用了一些新的技巧:
首先需要下載Tensorflow models Tensorflow models,以便使用其中的CIFAR-10數(shù)據(jù)的類.進(jìn)入目錄models/tutorials/image/cifar10目錄,執(zhí)行以下代碼
import cifar10 import cifar10_input import tensorflow as tf import numpy as np import time # 定義batch_size, 訓(xùn)練輪數(shù)max_steps, 以及下載CIFAR-10數(shù)據(jù)的默認(rèn)路徑 max_steps = 3000 batch_size = 128 data_dir = 'E:\\tmp\cifar10_data\cifar-10-batches-bin' # 定義初始化weight的函數(shù),定義的同時(shí),對(duì)weight加一個(gè)L2 loss,放在集'losses'中 def variable_with_weight_loss(shape, stddev, w1): var = tf.Variable(tf.truncated_normal(shape, stddev=stddev)) if w1 is not None: weight_loss = tf.multiply(tf.nn.l2_loss(var), w1, name='weight_loss') tf.add_to_collection('losses', weight_loss) return var # 使用cifar10類下載數(shù)據(jù)集,并解壓、展開(kāi)到其默認(rèn)位置 #cifar10.maybe_download_and_extract() # 在使用cifar10_input類中的distorted_inputs函數(shù)產(chǎn)生訓(xùn)練需要使用的數(shù)據(jù)。需要注意的是,返回的是已經(jīng)封裝好的tensor, # 且對(duì)數(shù)據(jù)進(jìn)行了Data Augmentation(水平翻轉(zhuǎn)、隨機(jī)剪切、設(shè)置隨機(jī)亮度和對(duì)比度、對(duì)數(shù)據(jù)進(jìn)行標(biāo)準(zhǔn)化) images_train, labels_train = cifar10_input.distorted_inputs(data_dir=data_dir, batch_size=batch_size) # 再使用cifar10_input.inputs函數(shù)生成測(cè)試數(shù)據(jù),這里不需要進(jìn)行太多處理 images_test, labels_test = cifar10_input.inputs(eval_data=True, data_dir=data_dir, batch_size=batch_size) # 創(chuàng)建數(shù)據(jù)的placeholder image_holder = tf.placeholder(tf.float32, [batch_size, 24, 24, 3]) label_holder = tf.placeholder(tf.int32, [batch_size]) # 創(chuàng)建第一個(gè)卷積層 weight1 = variable_with_weight_loss(shape=[5, 5, 3, 64], stddev=5e-2, w1=0.0) kernel1 = tf.nn.conv2d(image_holder, weight1, strides=[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') # LRN層對(duì)ReLU會(huì)比較有用,但不適合Sigmoid這種有固定邊界并且能抑制過(guò)大值的激活函數(shù) norm1 = tf.nn.lrn(pool1, 4, bias=1.0, alpha=0.001 / 9.0, beta=0.75) # 創(chuàng)建第二個(gè)卷積層 weight2 = variable_with_weight_loss(shape=[5, 5, 64, 64], stddev=5e-2, w1=0.0) kernel2 = tf.nn.conv2d(norm1, weight2, strides=[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') # 使用一個(gè)全連接層 reshape = tf.reshape(pool2, [batch_size, -1]) dim = reshape.get_shape()[1].value weight3 = variable_with_weight_loss(shape=[dim, 384], stddev=0.04, w1=0.004) bias3 = tf.Variable(tf.constant(0.1, shape=[384])) local3 = tf.nn.relu(tf.matmul(reshape, weight3) + bias3) # 再使用一個(gè)全連接層,隱含節(jié)點(diǎn)數(shù)下降了一半,只有192個(gè),其他的超參數(shù)保持不變 weight4 = variable_with_weight_loss(shape=[384, 192], stddev=0.04, w1=0.004) bias4 = tf.Variable(tf.constant(0.1, shape=[192])) local4 = tf.nn.relu(tf.matmul(local3, weight4) + bias4) # 最后一層,將softmax放在了計(jì)算loss部分 weight5 = variable_with_weight_loss(shape=[192, 10], stddev=1 / 192.0, w1=0.0) bias5 = tf.Variable(tf.constant(0.0, shape=[10])) logits = tf.add(tf.matmul(local4, weight5), bias5) # 定義loss 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') # 獲取最終的loss loss = loss(logits, label_holder) # 優(yōu)化器 train_op = tf.train.AdamOptimizer(1e-3).minimize(loss) # 使用tf.nn.in_top_k函數(shù)求輸出結(jié)果中top k的準(zhǔn)確率,默認(rèn)使用top 1,也就是輸出分?jǐn)?shù)最高的那一類的準(zhǔn)確率 top_k_op = tf.nn.in_top_k(logits, label_holder, 1) # 使用tf.InteractiveSession創(chuàng)建默認(rèn)的session,接著初始化全部模型參數(shù) sess = tf.InteractiveSession() tf.global_variables_initializer().run() # 啟動(dòng)圖片數(shù)據(jù)增強(qiáng)線程 tf.train.start_queue_runners() # 正式開(kāi)始訓(xùn)練 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_time if step % 10 == 0: example_per_sec = batch_size / duration sec_per_batch = float(duration) format_str = 'step %d, loss=%.2f ,%.1f examples/sec, %.3f sec/batch' print(format_str % (step, loss_value, example_per_sec, sec_per_batch)) num_examples = 10000 import math num_iter = int(math.ceil(num_examples / batch_size)) true_count = 0 total_sample_count = num_iter * batch_size step = 0 while 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_holder}) true_count += np.sum(predictions) step += 1 precision = true_count / total_sample_count print('precision @ 1 = %.3f'%precision)
新聞標(biāo)題:Tensorflow卷積神經(jīng)網(wǎng)絡(luò)實(shí)例進(jìn)階-創(chuàng)新互聯(lián)
當(dāng)前地址:http://chinadenli.net/article24/pesce.html
成都網(wǎng)站建設(shè)公司_創(chuàng)新互聯(lián),為您提供服務(wù)器托管、網(wǎng)站內(nèi)鏈、品牌網(wǎng)站設(shè)計(jì)、微信小程序、企業(yè)建站、做網(wǎng)站
聲明:本網(wǎng)站發(fā)布的內(nèi)容(圖片、視頻和文字)以用戶投稿、用戶轉(zhuǎn)載內(nèi)容為主,如果涉及侵權(quán)請(qǐng)盡快告知,我們將會(huì)在第一時(shí)間刪除。文章觀點(diǎn)不代表本網(wǎng)站立場(chǎng),如需處理請(qǐng)聯(lián)系客服。電話:028-86922220;郵箱:631063699@qq.com。內(nèi)容未經(jīng)允許不得轉(zhuǎn)載,或轉(zhuǎn)載時(shí)需注明來(lái)源: 創(chuàng)新互聯(lián)
猜你還喜歡下面的內(nèi)容