tensorboard 自编码分类网络和vgg19 网络结构可视化

原文链接: tensorboard 自编码分类网络和vgg19 网络结构可视化

上一篇: Python 使用http服务器 在局域网内分享文件

下一篇: mobilenet_v2_1.4_224 flowers 数据集分类网络

可视化网络结构，只需要执行下面代码即可

    summary_writer = tf.summary.FileWriter('./log/', sess.graph)

tf.train.Saver 用来保存和加载模型参数

saver = tf.train.Saver()
saver.save(sess, './log/save')
saver.restore(sess, CKPT_PATH)

会在指定目录下生成event文件，按照时间排序，会显示最新的event中的网络结构

在event文件夹中执行下面代码，如果event文件有更新会自动重新加载，只需要刷新浏览器即可

 tensorboard.exe --logdir=.

自编码分类网络可视化

可以看到第一个loss是自编码loss

第二个是分类网络的loss

import tensorflow as tf
import numpy as np
import tensorflow.contrib.slim as slim
import cv2 as cv

TRAIN_PATH = "D:/data/Digit Recognizer/train.csv"
TEST_PATH = "D:/data/Digit Recognizer/test.csv"
learning_rate = .0001
TRAIN_STEP = 30000
# 大样本
BATCH_SIZE = 512
TEST_SIZE = 512
SHOW_STEP = 100

# 增大特征数据量
IMAGE_SIZE = 32

in_x = tf.placeholder(tf.float32, (None, IMAGE_SIZE, IMAGE_SIZE, 1))
in_y = tf.placeholder(tf.float32, (None, IMAGE_SIZE, IMAGE_SIZE, 1))
in_y2 = tf.placeholder(tf.float32, (None, 10))

kernel = (5, 5)
# 自编码网络
# 浅一点的网络效果稍微好点。。。
# 层数比较多的话提取的信息时需要使用大样本128
# 导致后面全连接分类效果比较差，误差跳动很厉害
# relu 函数比relu6的效果好一点。。。
with slim.arg_scope(
        [slim.conv2d, slim.conv2d_transpose],
        activation_fn=tf.nn.relu,
        # activation_fn=tf.nn.relu6,
):
    with tf.variable_scope("encode"):
        print(in_x.shape)
        net = slim.conv2d(in_x, 32, kernel, stride=2)
        print(net.shape)
        net = slim.conv2d(net, 16, kernel, stride=2)
        print(net.shape)
        core = slim.conv2d(net, 8, kernel, stride=2)
        print(core.shape)
    with tf.variable_scope("decode"):
        net = slim.conv2d_transpose(core, 16, kernel, stride=2)
        print(net.shape)
        net = slim.conv2d_transpose(net, 32, kernel, stride=2)
        print(net.shape)
        out_y = slim.conv2d_transpose(net, 1, kernel, stride=2)
        print(out_y.shape)

# 自编码loss和train
with tf.variable_scope('loss1'):
    loss_code = tf.reduce_mean((out_y - in_y) ** 2)
with tf.variable_scope('train1'):
    train_code = tf.train.AdamOptimizer(learning_rate).minimize(loss_code)

# 分类网络
with tf.variable_scope("class"):
    with slim.arg_scope(
            [slim.fully_connected],
            activation_fn=tf.nn.relu,
    ):
        net2 = slim.flatten(core)
        print(net2.shape)
        net2 = slim.fully_connected(net2, 128)
        print(net2.shape)
        net2 = slim.fully_connected(net2, 64)
        print(net2.shape)
        net2 = slim.fully_connected(net2, 32)
        print(net2.shape)
        net2 = slim.fully_connected(net2, 10)
        print(net2.shape)

# 交叉熵效果更好
# 均方误差0.6953125
with tf.variable_scope('loss2'):
    loss2 = tf.reduce_mean((net2 - in_y2) ** 2)

# 交叉熵可以达到.98
# loss2 = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(labels=in_y2, logits=net2))
with tf.variable_scope('train2'):
    train2 = tf.train.AdamOptimizer(learning_rate).minimize(loss2)

with tf.variable_scope('predict'):
    predict = tf.equal(tf.argmax(net2, 1), tf.argmax(in_y2, 1))
with tf.variable_scope('accuracy'):
    accuracy = tf.reduce_mean(tf.cast(predict, tf.float32))

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    summary_writer = tf.summary.FileWriter('./log/', sess.graph)

展开节点查看具体网络结构

vgg19 网络可视化

下载slim模块，加载网络，可视化网络结构

from nets.vgg import vgg_19
import tensorflow as tf

in_x = tf.placeholder(tf.float32, (None, 224, 224, 3))
logits, end_points = vgg_19(in_x)

with tf.Session()as sess:
    sess.run(tf.global_variables_initializer())
    summary_writer = tf.summary.FileWriter('./log/', sess.graph)

可以看到各个节点的输出