一、模型基本知識
-
卷積神經網絡
-
卷積
-
採樣
二、模型定義
本文以CNN卷積神經網絡原型爲示例：
-
初始化輸入層，轉換成統一尺寸大小：對於本例子中全部resize成28*28大小
代碼塊
Python
input_layer = tf.reshape(features["x"], [-1, 28, 28, 1])
-
定義第一層，卷積層，卷積核大小和數量：卷積核數量32個，每個卷積核大小5✖️5
代碼塊
Python
conv1 = tf.layers.conv2d(
inputs=input_layer,
filters=32,
kernel_size=[5, 5],
padding="same",
activation=tf.nn.relu)
-
定義第二層，池化層，所採用的池化方法爲最大池化方法，池化的batch大小爲：2✖️2，步長爲2
代碼塊
Python
pool1 = tf.layers.max_pooling2d(inputs=conv1, pool_size=[2, 2], strides=2)
-
定義第三層，卷積層，卷積核大小和數量：卷積核數量64個，每個卷積核大小5✖️5
代碼塊
Python
conv2 = tf.layers.conv2d( inputs=pool1, filters=64, kernel_size=[5, 5], padding="same", activation=tf.nn.relu)
-
定義第四層，池化層，所採用的池化方法爲最大池化方法，池化的batch大小爲：2✖️2，步長爲2
代碼塊
Python
pool2 = tf.layers.max_pooling2d(inputs=conv2, pool_size=[2, 2], strides=2)
-
定義第五層，全連接層，將第四層降採樣之後的數據resize層7*7*64以爲向量大小
代碼塊
Python
# Dense Layer pool2_flat = tf.reshape(pool2, [-1, 7 * 7 * 64]) dense = tf.layers.dense(inputs=pool2_flat, units=1024, activation=tf.nn.relu) dropout = tf.layers.dropout( inputs=dense, rate=0.4, training=mode == tf.estimator.ModeKeys.TRAIN) # Logits Layer logits = tf.layers.dense(inputs=dropout, units=10)
-
定義第五層，利用softmax算法計算預測每個類別的概率：
---訓練樣本，更新損失函數，更新梯度
---測試樣本，返回預測結果值
三、訓練數據代碼
過程：讀取數據-》構建結果評估模型-》訓練樣本-》預測樣本
# -*- coding: UTF-8 -*-
import tensorflow as tf
import numpy as np
import cnn_mnist
import input_data
# Load training and eval data
#mnist = tf.contrib.learn.datasets.load_dataset("/Users/dingjie/PycharmProjects/baby-robot/minist/MNIST-data/")
#mnist = tf.contrib.learn.datasets.load_dataset("MNIST-data")
mnist = input_data.read_data_sets("MNIST_data/")
train_data = mnist.train.images # Returns np.array
train_labels = np.asarray(mnist.train.labels, dtype=np.int32)
eval_data = mnist.test.images # Returns np.array
eval_labels = np.asarray(mnist.test.labels, dtype=np.int32)
# Create the Estimator ---構建評估模型
mnist_classifier = tf.estimator.Estimator(
model_fn=cnn_mnist.cnn_model_fn, model_dir="model")
# Set up logging for predictions
tensors_to_log = {"probabilities": "softmax_tensor"}
logging_hook = tf.train.LoggingTensorHook(
tensors=tensors_to_log, every_n_iter=50)
# Train the model ---初始化訓練數據
train_input_fn = tf.estimator.inputs.numpy_input_fn(
x={"x": train_data},
y=train_labels,
batch_size=100,
num_epochs=None,
shuffle=True)
#訓練數據，得到訓練模型
mnist_classifier.train(
input_fn=train_input_fn,
steps=20000,
hooks=[logging_hook])
# Evaluate the model and print results ----結果評估
eval_input_fn = tf.estimator.inputs.numpy_input_fn(
x={"x": eval_data},
y=eval_labels,
num_epochs=1,
shuffle=False)
eval_results = mnist_classifier.evaluate(input_fn=eval_input_fn)
print(eval_results)