VGGNet,牛津大学计算机视觉组(Visual Geometry Group)和Google DeepMind公司一起研发,深度卷积神经网络。VGGNet反复堆叠3x3小型卷积核和2x2最大池化层,成功构筑16~19层深卷积神经网络。比state-of-the-art网络结构,错误率幅下降,取得ILSVRC 2014比赛分类第2名和定位第1名。拓展性强,迁移其他图片数据泛化性好。结构简洁,整个网络都用同样大小卷积核尺寸和最大池化尺寸。VGGNet训练后模型参数官方开源,domain specific图像分类任务再训练,提供较好初始化权重。
ConvNet Configuration A A-LRN B C D E weight layers 11 11 13 16 16 19 input(224x224 RGB image) conv3-64 conv3-64 conv3-64 conv3-64 conv3-64 conv3-64 LRN conv3-64 conv3-64 conv3-64 conv3-64 maxpool conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 maxpool conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv1-256 conv3-256 conv3-256 conv3-256 maxpool conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv1-512 conv3-512 conv3-512 conv3-512 maxpool conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv1-512 conv3-512 conv3-512 conv3-512 maxpool FC-4096 FC-4096 FC-1000 soft-max Network A,A-LRN B C D E Number of parameters 133 133 134 138 144
卷积层参数量少,最后3个全连接层参数多。训练耗时在卷积,计算量较大。D为VGGNet-16,E为VGGNet-19。C比B多3个1x1卷积层,线性变换,输入、输出通道数不变,没降维。
VGGNet 5段卷积,每段2~3卷积层,每段后接最大池化层给缩小图片尺寸。每段卷积核数量一样,越后段卷积核数量越多,64-128-256-512-512。多个3x3卷积层堆叠。2个3x3卷积层串联相当1个5x5。3个3x3卷积层串联相当1个7x7。 参数更少,非线性变换更多,增强特征学习能力。
先训练级别A简单网络,再复用A网络权重初如化复杂模型,训练收敛速度更快。预测,Multi-Scale,图像scale尺寸Q,图片输入卷积网络计算。最后卷积层,滑窗分类预测,不同窗口分类结果平均,不同尺寸Q结果平均得最后结果,提高图片数据利用率,提升预测准确率。训练过程,用Multi-Scale数据增强,原始图像缩放不同尺寸S,随机裁切224x224图片,增加数据量,防止过拟合。
LRN层作用不大,越深网络效果越好,1x1卷积很有效,但大卷积核可以学习更大空间特征。
载入系统库、TensorFlow。
conv_op函数,创建卷积层,参数存入参数列表。输入,input_op tensor,name 层名,kh kernel height 卷积核高,kw kernel width 卷积核宽,n_out 卷积核数量 输出通道数,dh 步长高,dw 步长宽,p参数列表。get_shape()[-1].value获取输入input_op通道数。tf.name_scope(name)设置scope。tf.get_variable创建kernel(卷积核),shape [kh,kw,n_in,n_out],卷积核高宽、输入输出通道数。tf.contrib.layers.xavier_initializer_conv2d()参数初始化。
tf.nn.conv2d卷积处理input_op。卷积核kernel,步长dhxdw,paddings模式SAME。tf.constant 赋值biases 0,tf.Variable转可训练参数。tf.nn.bias_add 相加卷积结果conv和bias,tf.nn.relu非线性处理得activation。创建卷积层,参数kernel、biases添加到参数列表p,卷积层输出activation返回。
全连接层创建函数 fc_op。先获取输入input_op通道数。tf.get_variable创建全连接层参数,第一维度输入通道数n_in,第二维度输出通道数n_out。xavier_initializer参数初始化。biases初始化0.1,避免dead neuron。tf.nn.relu_layer矩阵相乘input_op、kernel,加biases,ReLU非线性,交换得activation。全连接层参数kernel、biases添加参数列表p, activation返回。
定义最大池化层创建函数mpool_op。tf.nn.max_pool,输入input_op,池化尺寸khxkw,步长dhxdw,padding模式SAME。
VGGNet-16网络结构,6个部分,前5段卷积网络,最后一段全连接网络。定义创建VGGNet网络结构函数inference_op。输入input_op、keep_prob(控制dropout比率,placeholder)。先初始化参数列表p。
创建第一段卷积网络,两个卷积层(conv_op),一个最大池化层(mpool_op)。卷积核大小3x3,卷积核数量(输出通道数) 64,步长1x1,全像素扫描。第一卷积层输入input_op尺寸224x224x3,输出尺寸224x224x64。第二卷积层输入输出尺寸224x224x64。最大池化层2x2,输出112x112x64。
第二段卷积网络,2个卷积层,1个最大池化层。卷积输出通道数128。输出尺寸56x56x128。
第三段卷积网络,3个卷积层,1个最大池化层。卷积输出通道数256。输出尺寸28x28x256。
第四段卷积网络,3个卷积层,1个最大池化层。卷积输出通道数512。输出尺寸14x14x512。
第五段卷积网络,3个卷积层,1个最大池化层。卷积输出通道数512。输出尺寸7x7x512。输出结果每个样本,tf.reshape 扁平化为长度7x7x512=25088一维向量。
连接4096隐含点全连接层,激活函数ReLU。连接Dropout层,训练节点保留率0.5,预测1.0。
全连接层,Dropout层。
最后连接1000隐含点全连接层,Softmax 分类输出概率。tf.argmax 输出概率最大类别。返回fc8、softmax、predictions、参数列表p。
VGGNet-16网络结构构建完成。
评测函数time_tensorflow_run。session.run()方法,引入feed_dict,方便传入keep_prob控制Dropout层保留比率。
评测主函数run_benchmark。评测forward(inference)、backward(trainning)运算性能。生成尺寸224x224随机图片,tf.random_nornal函数生成标准差0.1正态分布随机数。
创建keep_prob placeholder,调用inference_op函数构建VGGNet-16网络结构,获得predictions、softmax、fc8、参数列表p。
创建Session,初始化全局参数。设keep_prob 1.0 预测。time_tensorflow_run评测forward运算时间。
计算VGGNet-16最后全连接层输出fc8 l2 loss。tf.gradients求loss所有模型参数梯度。time_tensorflow_run评测backward运算时间。target为求解梯度操作grad,keep_prob 0.5。设batch_size 32。
执行评测主函数run_benchmark(),测试VGGNet-16 TensorFlow forward、backward耗时。forward平均每个batch耗时0.152s。backward求解梯度,平均每个batch耗时0.617s。
VGGNet,7.3%错误率。更深网络,更小卷积核,隐式正则化。
from datetime import datetime import math import time import tensorflow as tf def conv_op(input_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 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, kernel, biases, name=scope) p += [kernel, biases] return activation def 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) def 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_out=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 56x56x128 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", 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) # 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, name="fc7_drop") fc8 = fc_op(fc7_drop, name="fc8", n_out=1000, p=p) softmax = tf.nn.softmax(fc8) predictions = tf.argmax(softmax, 1) return predictions, softmax, fc8, p def time_tensorflow_run(session, target, feed, info_string): num_steps_burn_in = 10 total_duration = 0.0 total_duration_squared = 0.0 for i in range(num_batches + num_steps_burn_in): start_time = time.time() _ = session.run(target, feed_dict=feed) duration = time.time() - start_time if i >= num_steps_burn_in: if not i % 10: print ('%s: step %d, duration = %.3f' % (datetime.now(), i - num_steps_burn_in, duration)) total_duration += duration total_duration_squared += duration * duration mn = total_duration / num_batches vr = total_duration_squared / num_batches - mn * mn sd = math.sqrt(vr) print ('%s: %s across %d steps, %.3f +/- %.3f sec / batch' % (datetime.now(), info_string, num_batches, mn, sd)) def run_benchmark(): with tf.Graph().as_default(): image_size = 224 images = tf.Variable(tf.random_normal([batch_size, image_size, image_size, 3], dtype=tf.float32, stddev=1e-1)) keep_prob = tf.placeholder(tf.float32) predictions, softmax, fc8, p = inference_op(images, keep_prob) init = tf.global_variables_initializer() config = tf.ConfigProto() config.gpu_options.allocator_type = 'BFC' sess = tf.Session(config=config) sess.run(init) time_tensorflow_run(sess, predictions, {keep_prob:1.0}, "Forward") objective = tf.nn.l2_loss(fc8) grad = tf.gradients(objective, p) time_tensorflow_run(sess, grad, {keep_prob:0.5}, "Forward-backward") batch_size=32 num_batches=100 run_benchmark()
参考资料:
《TensorFlow实战》
欢迎付费咨询(150元每小时),我的微信:qingxingfengzi