【VGG】复现VGG需要的函数

旧城等待， 2022-03-09 18:20 240阅读 0赞

# tf.placeholder() #

def placeholder(dtype, shape=None, name=None):

tf.placeholder()用来传入真是训练样本/测试/真实特征/待处理特征，仅占位，

不必给初值，用sess.run的feed\_dict参数以字典的形式喂入x

![20190313161855371.png][]

# no.save() / np.load() #

def save(file, arr, allow_pickle=True, fix_imports=True):

def load(file, mmap_mode=None, allow_pickle=True, fix_imports=True,
             encoding='ASCII'):

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3BsU29uZ19DU0RO_size_16_color_FFFFFF_t_70][]

使用示例

import numpy as np
    
    A = np.arange(15).reshape(3,5)
    print(A)
    np.save("A",A)
    B = np.load('A.npy')
    print(B)

运行结果

> \[\[ 0  1  2  3  4\]  
>  \[ 5  6  7  8  9\]  
>  \[10 11 12 13 14\]\]  
> \[\[ 0  1  2  3  4\]  
>  \[ 5  6  7  8  9\]  
>  \[10 11 12 13 14\]\]

# tf.shape() #

返回数据的维度

import tensorflow as tf
    import os
    import numpy as np
    
    os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"  # 忽略tensorflow警告信息
    
    x = tf.constant([[1,2,3],[4,5,6]])
    A = [[1,3,4],[4,5,6],[7,8,9]]
    y =np.arange(12).reshape(1,3,4)
    
    with tf.Session() as sess:
        print('A:',sess.run(tf.shape(A)))
        print('x:',sess.run(tf.shape(x)))
        print('y:',sess.run(tf.shape(y)))

结果：

> A: \[3 3\]  
> x: \[2 3\]  
> y: \[1 3 4\]

# 其他 #

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3BsU29uZ19DU0RO_size_16_color_FFFFFF_t_70 1][]

这里演示一下 tf.reshape()

import tensorflow as tf
    import os
    
    os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"  # 忽略tensorflow警告信息
    
    x = tf.constant([[1,2,3],[4,5,6]])
    
    a = tf.reshape(x,[2,-1]) # -1表示跟随另一列自动补全，自动计算
    b = tf.reshape(x,[-1,2])
    c = tf.reshape(x,[1,6])
    
    with tf.Session() as sess:
        print('a:',sess.run(tf.shape(a)))
        print('b:',sess.run(tf.shape(b)))
        print('c:',sess.run(tf.shape(c)))

运行结果

> a: \[2 3\]  
> b: \[3 2\]  
> c: \[1 6\]

[20190313161855371.png]: /images/20220309/8bdfa920a8c14b0995c2ddb6e7c4310b.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3BsU29uZ19DU0RO_size_16_color_FFFFFF_t_70]: /images/20220309/8560619922cd4b0a8938bd83d55c0f41.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3BsU29uZ19DU0RO_size_16_color_FFFFFF_t_70 1]: /images/20220309/eca7cb0288b74f309050bd4df18802d9.png