TenSorFlow学习笔记
只为敦促自己学习
看的见的时间
啥都不会的研究生
展开
-
十四.自定义层
import tensorflow as tffrom tensorflow import kerasclass Linear(keras.layers.Layer):#继承 def __init__(self,units=32,input_dim=32):#单元数,输入向量维数。初始化方法 super(Linear,self).__init__()# 权重# w_init = tf.random_normal_initializer()#标准原创 2021-10-28 21:13:57 · 288 阅读 · 0 评论 -
十三. FCN
1.语义分割图片import tensorflow as tfimport osimport numpy as npimport matplotlib.pyplot as plt# Press the green button in the gutter to run the script.if __name__ == '__main__': img = tf.io.read_file('trimaps/yorkshire_terrier_174.png') img = tf原创 2021-10-28 15:23:17 · 207 阅读 · 0 评论 -
十二. 图像定位
1.图像位置数据解析import tensorflow as tfimport matplotlib.pyplot as pltfrom matplotlib.patches import Rectanglefrom lxml import etree#lxml是python的一个解析库,支持HTML和XML的解析import numpyimport globif __name__ == '__main__': img = tf.io.read_file('./images/Abys原创 2021-10-26 18:13:37 · 351 阅读 · 0 评论 -
十一. 迁移学习
import tensorflow as tfimport glob as globdef load_process_image(path,label): image = tf.io.read_file(path) image = tf.image.decode_jpeg(image,channels=3) image = tf.image.resize(image,[256,256]) image = image/255 return image,label原创 2021-10-22 19:54:29 · 89 阅读 · 0 评论 -
十. Tensorboard
import tensorflow as tfimport datetimeimport os# Press the green button in the gutter to run the script.if __name__ == '__main__': (train_image, train_labels), (test_image, test_labels) = tf.keras.datasets.mnist.load_data() train_image = tf.e原创 2021-10-20 18:10:04 · 76 阅读 · 0 评论 -
九.微分与自定义训练
import tensorflow as tfif __name__ == '__main__': v = tf.Variable(0.0)#变量<tf.Variable 'Variable:0' shape=() dtype=float32, numpy=0.0> v.assign(5)#<tf.Variable 'UnreadVariable' shape=() dtype=float32, numpy=5.0> v.assign_add(1)#<原创 2021-10-19 20:02:55 · 100 阅读 · 0 评论 -
八. airplane,lake图像分类
import tensorflow as tfimport numpy as npimport glob#glob是python自带的一个操作文件的相关模块import randomdef load_image(path): image = tf.io.read_file(path) image = tf.image.decode_jpeg(image,channels=3)#jpg,png,gif....经过压缩编码,所以读取之后要解码还原成一个三维矩阵 image =原创 2021-10-17 07:46:48 · 137 阅读 · 0 评论 -
七.卷积神经网络
import tensorflow as tfimport matplotlib.pyplot as pltimport pandasimport numpy as npif __name__ == '__main__': (train_images,train_labels),(test_images,test_labels)=tf.keras.datasets.fashion_mnist.load_data() train_images = train_images/255原创 2021-10-15 14:34:43 · 109 阅读 · 0 评论 -
六. tf.dataset输入模块
文章目录一. from_tensor_slices二. shuffle,repeat,square三. 实例一. from_tensor_slicesimport tensorflow as tfif __name__ == '__main__': #把给定的元组、列表和张量等数据进行特征切片,张量 dataset = tf.data.Dataset.from_tensor_slices([1,2,3,4,5]) for ele in dataset: prin原创 2021-10-14 09:35:57 · 244 阅读 · 0 评论 -
五. Droupt抑制过拟合
过拟合Drouptmodel.add(tf.keras.layers.Flatten(input_shape=(28, 28))) model.add(tf.keras.layers.Dense(128, activation='relu')) model.add(tf.keras.layers.Dropout(0.5))#激活50%神经元 model.add(tf.keras.layers.Dense(128, activation='relu'))原创 2021-10-13 19:50:30 · 468 阅读 · 0 评论 -
四. softmax多分类
多个选项的分类问题,用softmax.softmax把神经网络输出值变为概率分布,要求每个样本必须属于某个类别,且所有类别均被覆盖。categorical_crossentropy和sparse_categorical_crossentropy都是计算多分类crossentropy的,只是对y的格式要求不同。1)如果是categorical_crossentropy,那y必须是one-hot处理过的2)如果是sparse_categorical_crossentropy,那y就是原始的整数形式,比原创 2021-10-13 15:52:11 · 629 阅读 · 1 评论 -
三. 逻辑回归与交叉熵
逻辑回归给出"是"或"否"分类问题使用交叉熵损失函数更有效binary_crossentropy计算二元交叉熵import tensorflow as tfimport pandas as pdimport matplotlib.pyplot as pltif __name__ == '__main__': data = pd.read_csv('dataset/credit-a.csv',header=None)#默认情况下,会把数据内容的第一行默认为字段名标题。header=Non原创 2021-10-13 15:06:00 · 138 阅读 · 0 评论 -
二. 多层感知器
import tensorflow as tfimport pandas as pdimport matplotlib.pyplot as pltif __name__ == '__main__': data = pd.read_csv('dataset/Advertising.csv') print(data.head())#显示前五行数据 """ TV radio newspaper sales TV,radio,newspaper预测sales原创 2021-10-12 20:03:44 · 71 阅读 · 0 评论 -
一. tf.keras线性回归
Model: "sequential"_________________________________________________________________Layer (type) Output Shape Param # =================================================================dense (Dense) (None, 1)原创 2021-10-12 19:12:24 · 147 阅读 · 0 评论