东方耀AI技术分享

标题: 理解高维矩阵的填充 [打印本页]
作者: 东方耀    时间: 2019-11-20 13:44
标题: 理解高维矩阵的填充
  1. # -*- coding: utf-8 -*-
  2. __author__ = u'东方耀 微信：dfy_88888'
  3. __date__ = '2019/11/20 8:14'
  4. __product__ = 'PyCharm'
  5. __filename__ = 'dfy_test_demo'

  7. import numpy as np
  8. # 理解高维矩阵的填充，向量可以直接赋值
  9. boxes_tensor = np.zeros((2, 3, 4, 4))
  10. print(boxes_tensor.shape)
  11. print('原始的模板矩阵：', boxes_tensor)
  12. # 模板tensor中最后有4个维度
  13. A = np.full(shape=(2, 3, 4), fill_value=10)
  14. B = np.full(shape=(2, 3, 4), fill_value=11)
  15. C = np.array([666]*4)
  16. CC = np.full(shape=(2, 3, 4), fill_value=666)
  17. D = np.array([888]*4)
  18. DD = np.full(shape=(2, 3, 4), fill_value=888)

  20. print(A)
  21. print(B)
  22. print(CC)
  23. print(DD)
  24. boxes_tensor[:, :, :, 0] = A
  25. print(boxes_tensor)
  26. boxes_tensor[:, :, :, 1] = B
  27. print(boxes_tensor)
  28. # boxes_tensor[:, :, :, 2] = CC # 效果相同
  29. boxes_tensor[:, :, :, 2] = C
  30. print(boxes_tensor)
  31. # boxes_tensor[:, :, :, 3] = DD # 效果相同
  32. boxes_tensor[:, :, :, 3] = D
  33. print(boxes_tensor)
复制代码

  1. # 类似于归一化的操作 Normalization
  2. boxes_tensor[:, :, :, [0, 1]] /= 13
  3. boxes_tensor[:, :, :, [2, 3]] /= 1000
  4. print(boxes_tensor)
  5. print('查看高维矩阵的具体值：')
  6. # boxes_tensor.shape (2, 3, 4, 4)
  7. print(boxes_tensor[0, 0, 0:4, :])
  8. print(boxes_tensor[1, 2, 0:4, :])
复制代码








欢迎光临 东方耀AI技术分享 (http://www.ai111.vip/) Powered by Discuz! X3.4