keras的dataloader实现分析2

这里用numpy实现了一个keras的dataloader一个小栗子，dataloader生成是一个元组----（img_batch，label_batch），

img_batch为（n,h,w,c）形状；下面生成一个batch-size为2的img_batch

# import keras
import numpy as np
import os

a=np.random.randint(0,3,(2,3),dtype=np.uint8)
b=np.random.randint(3,5,(2,3),dtype=np.uint8)

# c=list(a)
# print(c)

c=[[1,2,3],[1,2,3]]
d=[[3,4,5],[3,4,5]]
# c=np.array(c)
# d=np.array(d)
c=np.array(a)
d=np.array(b)

c=np.expand_dims(c,axis=2)
d=np.expand_dims(d,axis=2)
e=[]
e.append(c)
e.append(d)
e=np.array(e)
print(e.shape)
print(e)

输出：

(2, 2, 3, 1)
[[[[0]
   [0]
   [2]]

  [[1]
   [1]
   [1]]]


 [[[3]
   [4]
   [4]]

  [[3]
   [4]
   [3]]]]

参考：https://blog.csdn.net/w5688414/article/details/84593705?utm_medium=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-8.edu_weight&depth_1-utm_source=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-8.edu_weight

 

https://blog.csdn.net/weixin_37737254/article/details/103884255?utm_medium=distribute.pc_relevant_t0.none-task-blog-OPENSEARCH-1.edu_weight&depth_1-utm_source=distribute.pc_relevant_t0.none-task-blog-OPENSEARCH-1.edu_weight

 

https://blog.csdn.net/owenfy/article/details/85467330?utm_medium=distribute.pc_relevant.none-task-blog-OPENSEARCH-3.edu_weight&depth_1-utm_source=distribute.pc_relevant.none-task-blog-OPENSEARCH-3.edu_weight

 

https://blog.csdn.net/owenfy/article/details/85467330?utm_medium=distribute.pc_relevant.none-task-blog-baidulandingword-4&spm=1001.2101.3001.4242