python实现npy格式文件转换为txt文件操作

如下代码会将npy的格式数据读出，并且输出来到控制台：

import numpy as np

##设置全部数据，不输出省略号
import sys
np.set_printoptions(threshold=sys.maxsize)

boxes=np.load('./input_output/boxes.npy')
print(boxes)
np.savetxt('./input_output/boxes.txt',boxes,fmt='％s',newline='\n')
print('---------------------boxes--------------------------')

如下代码实现npy格式文件转换为txt,并且保存到当前目录相同文件名

实现转换整个文件夹下面多个文件：

import os
import numpy as np
path='./input_output' #一个文件夹下多个npy文件，
txtpath='./input_output'
namelist=[x for x in os.listdir(path)]
for i in range( len(namelist) ):
datapath=os.path.join(path,namelist[i]) #specific address
print(namelist[i])
data = np.load(datapath).reshape([-1, 2]) # (39, 2)
np.savetxt('％s/％s.txt'％(txtpath,namelist[i]),data)
print ('over')
import os
import numpy as np
path='./input_output' #一个文件夹下多个npy文件
txtpath='./input_output'
namelist=[x for x in os.listdir(path)]
for i in range( len(namelist) ):
datapath=os.path.join(path,namelist[i]) #specific address
print(namelist[i])
#data = np.load(datapath).reshape([-1, 2]) # (39, 2)
input_data = np.load(datapath) # (39, 2)
data = input_data.reshape(1, -1)
np.savetxt('％s/％s.txt'％(txtpath,namelist[i]),data)
print ('over')

同样的代码，实现读取单个npy文件，读取并且存储为txt :

import numpy as np
input_data = np.load(r"C:\test.npy")
print(input_data.shape)
data = input_data.reshape(1,-1)
print(data.shape)
print(data)
np.savetxt(r"C:\test.txt",data,delimiter=',')

修改pycharm的控制台的buffer大小：

如果你是用pycharm作为Python的编辑器，那么控制台的buf默认为1024，如果输出数据太多，需要修改buff大小才能让

全部数据输出，修改方法：

找到 pycharm 安装目录的 bin 目录下 idea.properties 文件， 修改 idea.cycle.buffer 值，原来默认为 1024

#--------------------------------------------------------------------- # This option controls console cyclic buffer: keeps the console output size not higher than the specified buffer size (Kb). # Older lines are deleted. In order to disable cycle buffer use idea.cycle.buffer.size=disabled #--------------------------------------------------------------------- idea.cycle.buffer.size=102400

补充知识：读取npy格式的文件

npy文件保存的是网络的权重

问题：Ubuntu环境下用gedit打开npy文件，是这样的，根本看不了内容

解决方法：编写如下代码，使解码后的文件内容输出在控制台

import numpy as np
context = np.load('E:/KittiSeg_pretrained0/vgg16.npy',encoding="latin1")
print(context)

文件位置依据自己的存放位置进行修改

运行代码输出结果为

{'conv1_2': [array([[[[ 1.66219279e-01, 1.42701820e-01, -4.02113283e-03, ...,
     6.18828237e-02, -1.74057148e-02, -3.00644431e-02],
    [ 9.46945231e-03, 3.87477316e-03, 5.08365929e-02, ...,
    -2.77981739e-02, 1.71373668e-03, 6.82722731e-03],
    [ 6.32681847e-02, 2.12877709e-02, -1.63465310e-02, ...,
     8.80054955e-04, 6.68104272e-03, -1.41139806e-03],
    ...,
    [ 3.47490981e-03, 8.47019628e-02, -4.07223180e-02, ...,
    -1.13523193e-02, -7.48998486e-03, 3.19077494e-03],
    [ 5.97234145e-02, 4.97663505e-02, -3.23118735e-03, ...,
     1.43114366e-02, 3.03175431e-02, -4.23925705e-02],
    [ 1.33459672e-01, 4.95484173e-02, -1.78808011e-02, ...,
     2.25385167e-02, 3.02020740e-02, -2.17075031e-02]],

[[ 2.12007999e-01, 2.10127644e-02, -1.47626130e-02, ...,
     2.29580477e-02, 1.23102348e-02, -3.08422819e-02],
    [-2.62175221e-03, 7.42094172e-03, 6.74030930e-02, ...,
    -3.06594316e-02, 1.80578313e-03, 4.27369215e-03],
    [ 2.27197763e-02, -1.07841045e-02, -1.31095545e-02, ...,
    -1.15751950e-02, 4.18359675e-02, -1.92268589e-03],
    ...,
    [-2.70304317e-03, 7.41161704e-02, -3.32262330e-02, ...,
    -1.10277236e-02, 1.39831286e-02, 5.34419343e-03],
    [-3.20506282e-02, -2.40584910e-02, -4.52397857e-03, ...,
    -6.04042644e-03, 2.01962605e-01, -5.04491515e-02],
    [ 1.68114193e-02, -2.33167298e-02, -1.40886130e-02, ...,
    -7.79278344e-03, 1.28428593e-01, -2.58184522e-02]],

[[-5.91698708e-03, -2.26223674e-02, 4.88128467e-03, ...,
   4.13784146e-04, -4.84175496e-02, 1.63675251e-03],
  [-3.93767562e-03, 9.07397643e-03, 5.36517277e-02, ...,
  -2.56106984e-02, -4.17886395e-03, 2.47476017e-03],
  [-3.07008922e-02, -1.09781921e-02, -3.69096454e-03, ...,
  -1.19221993e-02, -1.39777903e-02, 8.52933805e-03],
  ...,
  ..........................................

来源：https://blog.csdn.net/liqi849478873/article/details/92807316