导入模块:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
from tensorflow.keras.layers import GaussianNoise
from tensorflow.keras.datasets import mnist
(X_train, y_train), (X_test, y_test) = mnist.load_data()
重新缩放数据
X_train = X_train/255
X_test = X_test/255
plt.imshow(X_train[0])
使用std dev = 0.2添加高斯噪声
sample = GaussianNoise(0.2)
noisey = sample(X_test[0:2],training=True) #plt.imshow(noisey[0])
获取错误:
ValueError: Tensor conversion requested dtype float64 for Tensor with dtype float32: 'Tensor("gaussian_noise_4_1/random_normal:0", shape=(2, 28, 28), dtype=float32)'
答案 0 :(得分:2)
类型转换非常昂贵,因此Tensorflow不会执行自动类型转换。默认情况下,Tensorflow的dtype为float32
,而导入的数据集的dtype为float64
。您只需将可选的dtype参数传递给GaussianNoise
:
sample = GaussianNoise(0.2, dtype=tf.float64)
或强制转换数组:
noisey = sample(X_test[0:2].astype(np.float32),training=True)
您可以通过许多不同的方式执行此操作:重要的是它们必须协调一致。
答案 1 :(得分:0)