|
- # -*- coding: utf-8 -*-
- __author__ = u'东方耀 微信:dfy_88888'
- __date__ = '2019/11/20 8:14'
- __product__ = 'PyCharm'
- __filename__ = 'dfy_test_demo'
- import numpy as np
- # 理解高维矩阵的填充,向量可以直接赋值
- boxes_tensor = np.zeros((2, 3, 4, 4))
- print(boxes_tensor.shape)
- print('原始的模板矩阵:', boxes_tensor)
- # 模板tensor中最后有4个维度
- A = np.full(shape=(2, 3, 4), fill_value=10)
- B = np.full(shape=(2, 3, 4), fill_value=11)
- C = np.array([666]*4)
- CC = np.full(shape=(2, 3, 4), fill_value=666)
- D = np.array([888]*4)
- DD = np.full(shape=(2, 3, 4), fill_value=888)
- print(A)
- print(B)
- print(CC)
- print(DD)
- boxes_tensor[:, :, :, 0] = A
- print(boxes_tensor)
- boxes_tensor[:, :, :, 1] = B
- print(boxes_tensor)
- # boxes_tensor[:, :, :, 2] = CC # 效果相同
- boxes_tensor[:, :, :, 2] = C
- print(boxes_tensor)
- # boxes_tensor[:, :, :, 3] = DD # 效果相同
- boxes_tensor[:, :, :, 3] = D
- print(boxes_tensor)
复制代码
- # 类似于归一化的操作 Normalization
- boxes_tensor[:, :, :, [0, 1]] /= 13
- boxes_tensor[:, :, :, [2, 3]] /= 1000
- print(boxes_tensor)
- print('查看高维矩阵的具体值:')
- # boxes_tensor.shape (2, 3, 4, 4)
- print(boxes_tensor[0, 0, 0:4, :])
- print(boxes_tensor[1, 2, 0:4, :])
复制代码
|
|