numpy和keras的resize顺序问题

最新推荐文章于 2024-05-21 08:00:00 发布

yxpandjay

最新推荐文章于 2024-05-21 08:00:00 发布

阅读量1k

点赞数

分类专栏： Python

本文链接：https://blog.csdn.net/yxpandjay/article/details/99407979

版权

Python 专栏收录该内容

26 篇文章 1 订阅

订阅专栏

先看实验代码:

import numpy as np
from keras.models import Sequential
from keras.layers import Reshape

def GetModel():
    model = Sequential()
    model.add(Reshape( (-1,), input_shape=(2,3,4)))
    return model

if __name__ == "__main__":
    r=np.random.randint(0,100,size=(2,3,4))
    reshape_size = (-1,4)
    r2=r.reshape(reshape_size,order='C')
    r3 = r.reshape( (-1)  )
    print("r",r)
    print("*************************")
    print("r2",r2)
    print("*************************")
    print("r3",r3)

    my_input = np.expand_dims(r,axis=0)
    model = GetModel()
    result = model.predict(my_input)
    print("*************************")
    print("result",result)



结果:
r [[[22 96  4 45]
  [38 39 12 82]
  [45 47 73 24]]

 [[18 95 26 84]
  [44 44 71 60]
  [47 19 20 77]]]
*************************
r2 [[22 96  4 45]
 [38 39 12 82]
 [45 47 73 24]
 [18 95 26 84]
 [44 44 71 60]
 [47 19 20 77]]
*************************
r3 [22 96  4 45 38 39 12 82 45 47 73 24 18 95 26 84 44 44 71 60 47 19 20 77]
*************************
result [[22. 96.  4. 45. 38. 39. 12. 82. 45. 47. 73. 24. 18. 95. 26. 84. 44. 44. 71. 60. 47. 19. 20. 77.]]

在结合这篇文章:https://blog.csdn.net/zhanggonglalala/article/details/79356653

简单来说，默认的order='C',也就是后面的维度变化快，前面的维度变化最慢，按照这个顺序去reshape

再附上一个实验代码:

import os
import tensorflow as tf
from keras.models import Model
from keras.optimizers import SGD,Adam
from keras.callbacks import ModelCheckpoint,TensorBoard
from keras.layers import *
from keras import backend as K
from keras.layers.normalization import BatchNormalization
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf 
from keras.utils import plot_model

def GetModel():

    inputs = Input(shape=(3,3,2))
    nets = Reshape(target_shape=(-1,1) )(inputs)

    model = Model(inputs=inputs,outputs=nets)
    return model

if __name__ == "__main__":
    model = GetModel()
    mat = np.random.randint(1,100,size=(3,3,2))
    print(mat)
    print("mat shape:",mat.shape)
    print(mat[0,1,:])
    mat = np.asarray(mat)
    mat = np.expand_dims(mat,axis=0)
    print('(*******************)')
    predict = model.predict(mat)
    print(predict)
    print('predict shape',predict.shape)


-----------------------------------------------------------------
-----------------------------------------------------------------
Result:

[[[ 5 27]
  [62 98]
  [10 70]]

 [[15 54]
  [48 93]
  [10 64]]

 [[24 12]
  [61 99]
  [53 57]]]
mat shape: (3, 3, 2)
[62 98]
(*******************)
[[[ 5.]
  [27.]
  [62.]
  [98.]
  [10.]
  [70.]
  [15.]
  [54.]
  [48.]
  [93.]
  [10.]
  [64.]
  [24.]
  [12.]
  [61.]
  [99.]
  [53.]
  [57.]]]
predict shape (1, 18, 1)