语义分割入门之路

编程入门 行业动态 更新时间:2024-10-13 20:18:19

<a href=https://www.elefans.com/category/jswz/34/1769519.html style=语义分割入门之路"/>

语义分割入门之路

语义分割入门之路_1,理论部分_1吴恩达的深度学习课程

  • 一,第一门课 神经网络和深度学习(Neural Networks and Deep Learning)
    • 1,第一周:深度学习引言(Introduction to Deep Learning)
      • 1.1 欢迎(Welcome)
      • 1.2 什么是神经网络?(What is a Neural Network)
      • 1.3 神经网络的监督学习(Supervised Learning with Neural Networks)
      • 1.4 为什么深度学习会兴起?(Why is Deep Learning taking off?)
    • 2,第二周:神经网络的编程基础(Basics of Neural Network programming)
      • 2.1 二分类(Binary Classification)
      • 2.2逻辑回归(Logistic Regression)
      • 2.3逻辑回归的代价函数(Logistic Regression Cost Function)
      • 2.4梯度下降法(Gradient Descent)
      • 2.5导数(Derivatives)
      • 2.6更多的导数例子(More Derivative Examples)
      • 2.7计算图(Computation Graph)
      • 2.8使用计算图求导数(Derivatives with a Computation Graph)
      • 2.9逻辑回归中的梯度下降(Logistic Regression Gradient Descent)
      • 2.10 m 个样本的梯度下降(Gradient Descent on m Examples)
      • 2.11向量化(Vectorization)
      • 2.12向量化的更多例子(More Examples of Vectorization)
      • 2.13向量化逻辑回归(Vectorizing Logistic Regression)
      • 2.14向量化 logistic 回归的梯度输出(Vectorizing Logistic Regression's Gradient)
      • 2.15Python 中的广播(Broadcasting in Python)
      • 2.16关于 python _ numpy 向量的说明(A note on python or numpy vectors)参考视频:
      • 2.17 Jupyter/iPython Notebooks快速入门(Quick tour of Jupyter/iPython Notebooks)
      • 2.18logistic 损失函数的解释(Explanation of logistic regression cost function)
    • 3,第三周:浅层神经网络(Shallow neural networks)
      • 3.1神经网络概述(Neural Network Overview)
      • 3.2神经网络的表示(Neural Network Representation)
      • 3.3计算一个神经网络的输出(Computing a Neural Network's output)
      • 3.4多样本向量化(Vectorizing across multiple examples)
      • 3.5向量化实现的解释(Justification for vectorized implementation)
      • 3.6 激活函数(Activation functions)
      • 3.7为什么需要非线性激活函数?(why need a nonlinear activation function?)
      • 3.8激活函数的导数(Derivatives of activation functions)
      • 3.9神经网络的梯度下降(Gradient descent for neural networks)
      • 3.10 选修)直观理解反向传(Backpropagation intuition)
      • 3.11随机初始化(Random Initialization)
    • 4,第四周:深层神经网络(Deep Neural Networks)
      • 4.1深层神经网络(Deep L-layer neural network)
      • 4.2前向传播和反向传播(Forward and backward propagation)
      • 4.3深层网络中的前向传播(Forward propagation in a Deep Network)
      • 4.4核对矩阵的维数(Getting your matrix dimensions right)
      • 4.5为什么使用深层表示?(Why deep representations?)
      • 4.6搭建神经网络块(Building blocks of deep neural networks)
      • 4.7参数VS超参数(Parameters vs Hyperparameters)
      • 4.8深度学习和大脑的关联性(What does this have to do with the brain?)
  • 二,第二门课 改善深层神经网络:超参数调试、正则化以及优化(Improving Deep Neural Networks:Hyperparameter tuning, Regularization and Optimization)
    • 1,第一周:深度学习的实用层面(Practical aspects of Deep Learning)
      • 1.1 训练,验证,测试集(Train / Dev / Test sets)
      • 1.2偏差,方差(Bias /Variance)
      • 1.3 机器学习基础(Basic Recipe for Machine Learning)
      • 1.4正则化(Regularization)
      • 1.5为什么正则化有利于预防过拟合呢?(Why regularization reduces overfitting?)
      • 1.6dropout 正则化(Dropout Regularization)
      • 1.7理解 dropout(Understanding Dropout)
      • 1.8其他正则化方法(Other regularization methods)
      • 1.9归一化输入(Normalizing inputs)
      • 1.10梯度消失/梯度爆炸(Vanishing / Exploding gradients)
      • 1.11神经网络的权重初始化(Weight Initialization for Deep NetworksVanishing / Exploding gradients)
      • 1.12梯度的数值逼近(Numerical approximation of gradients)
      • 1.13梯度检验(Gradient checking)
      • 1.14梯度检验应用的注意事项(Gradient Checking Implementation Notes)
    • 2,第二周:优化算法 (Optimization algorithms)
      • 2.1 Mini-batch 梯度下降(Mini-batch gradient descent)
      • 2.2理解mini-batch梯度下降法(Understanding mini-batch gradient descent)
      • 2.3指数加权平均数(Exponentially weighted averages)
      • 2.4理解指数加权平均数(Understanding exponentially weighted averages)
      • 2.5指数加权平均的偏差修正(Bias correction in exponentially weighted averages)
      • 2.6动量梯度下降法(Gradient descent with Momentum)
      • 2.7RMSprop
      • 2.8Adam 优化算法(Adam optimization algorithm)
      • 2.9学习率衰减(Learning rate decay)
      • 2.10局部最优的问题(The problem of local optima)
    • 3,第三周超参数调试,batch正则化和程序框架(Hyperparameter tuning, Batch Normalization and Programming Frameworks)
      • 3.1调试处理(Tuning process)
      • 3.2为超参数选择合适的范围(Using an appropriate scale to pick hyperparameters)
      • 3.3超参数调试的实践:Pandas VS Caviar(Hyperparameters tuning in practice: Pandas vs. Caviar)
      • 3.4归一化网络的激活函数(Normalizing activations in a network)
      • 3.5将 Batch Norm 拟合进神经网络(Fitting Batch Norm into a neural network)
      • 3.6Batch Norm 为什么奏效?(Why does Batch Norm work?)
      • 3.7测试时的 Batch Norm(Batch Norm at test time)
      • 3.8Softmax 回归(Softmax regression)
      • 3.9训练一个 Softmax 分类器(Training a Softmax classifier)
      • 3.10深度学习框架(Deep Learning frameworks)
      • 3.11TensorFlow
  • 三,第三门课 结构化机器学习项目 (Structuring Machine Learning Projects)
    • 1,第一周:机器学习策略(1)(ML Strategy (1))
      • 1.1为什么是ML策略?(Why ML Strategy?)
      • 1.2正交化(Orthogonalization)
      • 1.3单一数字评估指标(Single number evaluation metric)
      • 1.4满足和优化指标(Satisficing and optimizing metrics)
      • 1.5训练集、开发集、测试集的划分(Train/dev/test distributions)
      • 1.6开发集和测试集的大小 (Size of the dev and test sets)
      • 1.7什么时候改变开发集/测试集和评估指标(When to change dev/test sets and metrics)
      • 1.8为什么是人的表现 (Why human-level performance?)(未看)
      • 1.9可避免偏差(Avoidable bias)(未看)
      • 1.10 理解人类的表现 (Understanding human-level performance)(未看)
      • 1.11 超过人类的表现(Surpassing human-level performance)(未看)
      • 1.12改善你的模型表现 (Improving your model performance)(未看)
    • 2,第二周:机器学习策略(2)(ML Strategy (2))(未看)
  • 四,第四门课 卷积神经网络(Convolutional Neural Networks)
    • 1,第一周 卷积神经网络(Foundations of Convolutional Neural Networks)
      • 1.1计算机视觉(Computer vision)
      • 1.2边缘检测示例(Edge detection example)
      • 1.3更多边缘检测内容(More edge detection)
      • 1.4填充(Padding)
      • 1.5 卷积步长(Strided convolutions)
      • 1.6三维卷积(Convolutions over volumes)
      • 1.7单层卷积网络(One layer of a convolutional network)
      • 1.8简单卷积网络示例(A simple convolution network example)
      • 1.9 池化层(Pooling layers)
      • 1.10卷积神经网络示例(Convolutional neural network example)
      • 1.11为什么使用卷积?(Why convolutions?)
    • 2,第二周 深度卷积网络:实例探究(Deep convolutional models: case studies)
      • 2.1为什么要进行实例探究?(Why look at case studies?)
      • 2.2经典网络(Classic networks)
      • 2.3残差网络(Residual Networks (ResNets))
      • 2.4残差网络为什么有用?(Why ResNets work?)
      • 2.5网络中的网络以及 1×1 卷积(Network in Network and 1×1 convolutions)
      • 2.6谷歌 Inception 网络简介(Inception network motivation)
      • 2.7Inception 网络(Inception network)
      • 2.8使用开源的实现方案(Using open-source implementations)
      • 2.9迁移学习(Transfer Learning)
      • 2.10数据扩充(Data augmentation)
      • 2.11计算机视觉现状(The state of computer vision)
    • 3,第三周 目标检测(Object detection) (未看)
    • 4,第四周 特殊应用:人脸识别和神经风格转换(Special applications: Face recognition &Neural style transfer)(未看)
  • 五,第五门课 序列模型(Sequence Models)(未看)

deeplearning.ai(吴恩达老师的深度学习课程笔记及资源)整理的电子版文档:/

一,第一门课 神经网络和深度学习(Neural Networks and Deep Learning)

1,第一周:深度学习引言(Introduction to Deep Learning)

1.1 欢迎(Welcome)

1.2 什么是神经网络?(What is a Neural Network)

1.3 神经网络的监督学习(Supervised Learning with Neural Networks)

1.4 为什么深度学习会兴起?(Why is Deep Learning taking off?)

2,第二周:神经网络的编程基础(Basics of Neural Network programming)

2.1 二分类(Binary Classification)

2.2逻辑回归(Logistic Regression)

2.3逻辑回归的代价函数(Logistic Regression Cost Function)

2.4梯度下降法(Gradient Descent)

2.5导数(Derivatives)

2.6更多的导数例子(More Derivative Examples)

2.7计算图(Computation Graph)

2.8使用计算图求导数(Derivatives with a Computation Graph)

2.9逻辑回归中的梯度下降(Logistic Regression Gradient Descent)

2.10 m 个样本的梯度下降(Gradient Descent on m Examples)

2.11向量化(Vectorization)

2.12向量化的更多例子(More Examples of Vectorization)

2.13向量化逻辑回归(Vectorizing Logistic Regression)

2.14向量化 logistic 回归的梯度输出(Vectorizing Logistic Regression’s Gradient)

2.15Python 中的广播(Broadcasting in Python)

2.16关于 python _ numpy 向量的说明(A note on python or numpy vectors)参考视频:

2.17 Jupyter/iPython Notebooks快速入门(Quick tour of Jupyter/iPython Notebooks)

2.18logistic 损失函数的解释(Explanation of logistic regression cost function)

3,第三周:浅层神经网络(Shallow neural networks)

3.1神经网络概述(Neural Network Overview)

3.2神经网络的表示(Neural Network Representation)

3.3计算一个神经网络的输出(Computing a Neural Network’s output)

3.4多样本向量化(Vectorizing across multiple examples)

3.5向量化实现的解释(Justification for vectorized implementation)

3.6 激活函数(Activation functions)

3.7为什么需要非线性激活函数?(why need a nonlinear activation function?)

3.8激活函数的导数(Derivatives of activation functions)

3.9神经网络的梯度下降(Gradient descent for neural networks)

3.10 选修)直观理解反向传(Backpropagation intuition)

3.11随机初始化(Random Initialization)

4,第四周:深层神经网络(Deep Neural Networks)

4.1深层神经网络(Deep L-layer neural network)

4.2前向传播和反向传播(Forward and backward propagation)

4.3深层网络中的前向传播(Forward propagation in a Deep Network)

4.4核对矩阵的维数(Getting your matrix dimensions right)

4.5为什么使用深层表示?(Why deep representations?)

4.6搭建神经网络块(Building blocks of deep neural networks)

4.7参数VS超参数(Parameters vs Hyperparameters)

4.8深度学习和大脑的关联性(What does this have to do with the brain?)

二,第二门课 改善深层神经网络:超参数调试、正则化以及优化(Improving Deep Neural Networks:Hyperparameter tuning, Regularization and Optimization)

1,第一周:深度学习的实用层面(Practical aspects of Deep Learning)

1.1 训练,验证,测试集(Train / Dev / Test sets)

1.2偏差,方差(Bias /Variance)

可另外参考:

1.3 机器学习基础(Basic Recipe for Machine Learning)

1.4正则化(Regularization)

1.5为什么正则化有利于预防过拟合呢?(Why regularization reduces overfitting?)

1.6dropout 正则化(Dropout Regularization)

1.7理解 dropout(Understanding Dropout)

1.8其他正则化方法(Other regularization methods)

1.9归一化输入(Normalizing inputs)

1.10梯度消失/梯度爆炸(Vanishing / Exploding gradients)

1.11神经网络的权重初始化(Weight Initialization for Deep NetworksVanishing / Exploding gradients)

1.12梯度的数值逼近(Numerical approximation of gradients)

1.13梯度检验(Gradient checking)

1.14梯度检验应用的注意事项(Gradient Checking Implementation Notes)

2,第二周:优化算法 (Optimization algorithms)

2.1 Mini-batch 梯度下降(Mini-batch gradient descent)

2.2理解mini-batch梯度下降法(Understanding mini-batch gradient descent)

2.3指数加权平均数(Exponentially weighted averages)

2.4理解指数加权平均数(Understanding exponentially weighted averages)

2.5指数加权平均的偏差修正(Bias correction in exponentially weighted averages)

2.6动量梯度下降法(Gradient descent with Momentum)

2.7RMSprop

2.8Adam 优化算法(Adam optimization algorithm)

2.9学习率衰减(Learning rate decay)

2.10局部最优的问题(The problem of local optima)

3,第三周超参数调试,batch正则化和程序框架(Hyperparameter tuning, Batch Normalization and Programming Frameworks)

3.1调试处理(Tuning process)

3.2为超参数选择合适的范围(Using an appropriate scale to pick hyperparameters)

3.3超参数调试的实践:Pandas VS Caviar(Hyperparameters tuning in practice: Pandas vs. Caviar)

3.4归一化网络的激活函数(Normalizing activations in a network)

3.5将 Batch Norm 拟合进神经网络(Fitting Batch Norm into a neural network)

3.6Batch Norm 为什么奏效?(Why does Batch Norm work?)

3.7测试时的 Batch Norm(Batch Norm at test time)

3.8Softmax 回归(Softmax regression)

3.9训练一个 Softmax 分类器(Training a Softmax classifier)

3.10深度学习框架(Deep Learning frameworks)

3.11TensorFlow

假设有一个损失函数需要最小化,Jw=w²-10w+25

# exercise
# 假设有一个损失函数需要最小化,Jw=w²-10w+25
#导入TensorFlow
import numpy as np
import tensorflow as tf#接下来,让我们定义参数w,在TensorFlow中,你要用tf.Variable()来定义参数
w = tf.Variable(0,dtype = tf.float32)#然后我们定义损失函数J:
# cost = tf.add(tf.add(w**2,tf.multiply(- 10.,w)),25)
cost = w**2-10*w+25#让我们用0.01的学习率,目标是最小化损失
train = tf.train.GradientDescentOptimizer(0.01).minimize(cost)#最后下面的几行是惯用表达式:
init = tf.global_variables_initializer()
session = tf.Session()#这样就开启了一个TensorFlow session。
session.run(init)#来初始化全局变量。#然后让TensorFlow评估一个变量,我们要用到:
session.run(w)
#上面的这一行将w初始化为0,并定义损失函数,我们定义train为学习算法,它用梯度下降法
#优化器使损失函数最小化,但实际上我们还没有运行学习算法,所以session.run(w)评估了w
print(session.run(w))session.run(train) #它所做的就是运行一步梯度下降法。#接下来在运行了一步梯度下降法后,让我们评估一下w的值,再print:
print(session.run(w))#在一步梯度下降法之后,w现在是0.1。
# 0.0
# 0.099999994session.run(train)
print(session.run(w))
# 0.29404for i in range(1000):session.run(train)
print(session.run(w))
# 4.999988

如何把训练数据加入TensorFlow程序呢?定义x,把它想做扮演训练数据的角色,训练数据有x和y,但这个例子中只有x,把x定义为:

coefficients = np.array([[1.],[-10.],[25.]])w=tf.Variable(0,dtype=tf.float32)
x=tf.placeholder(tf.float32,[3,1])
# cost = tf.add(tf.add(w**2,tf.multiply(- 10.,w)),25)
# cost = w**2-10*w+25
cost = x[0][0]*w**2+x[1][0]*w+x[2][0]
train = tf.train.GradientDescentOptimizer(0.01).minimize(cost)init = tf.global_variables_initializer()
session=tf.Session()
session.run(init)
print(session.run(w))
# 0.0session.run(train,feed_dict={x:coefficients})
print(session.run(w))
# 0.099999994for i in range(1000):session.run(train,feed_dict={x:coefficients})
print(session.run(w))
# 4.999988
coefficients = np.array([[1.],[-20.],[100.]])w=tf.Variable(0,dtype=tf.float32)
x=tf.placeholder(tf.float32,[3,1])
# cost = tf.add(tf.add(w**2,tf.multiply(- 10.,w)),25)
# cost = w**2-10*w+25
cost = x[0][0]*w**2+x[1][0]*w+x[2][0]
train = tf.train.GradientDescentOptimizer(0.01).minimize(cost)init = tf.global_variables_initializer()
session=tf.Session()
session.run(init)
print(session.run(w))session.run(train,feed_dict={x:coefficients})
print(session.run(w))for i in range(1000):session.run(train,feed_dict={x:coefficients})
print(session.run(w))# 0.0
# 0.19999999
# 9.999976
import numpy as np
import tensorflow as tfcoefficients=np.array([[1],[-20],[25]])w = tf.Variable([0],dtype=tf.float32)
x = tf.placeholder(tf.float32,[3,1])
cost=x[0][0]*w**2+x[1][0]*w+x[2][0]  #(w-5)**2
train=tf.train.GradientDescentOptimizer(0.01).minimize(cost)
init=tf.global_variables_initializer()
session=tf.Session()
session.run(init)
print(session.run(w))# with tf.Session() as session:
#     session.run(init)
#     print(session.run(w))for i in range(1000):session.run(train,feed_dict={x:coefficients})
print(session.run(w))
# [0.]
# [9.999976]

三,第三门课 结构化机器学习项目 (Structuring Machine Learning Projects)

1,第一周:机器学习策略(1)(ML Strategy (1))

1.1为什么是ML策略?(Why ML Strategy?)

1.2正交化(Orthogonalization)

1.3单一数字评估指标(Single number evaluation metric)

1.4满足和优化指标(Satisficing and optimizing metrics)

1.5训练集、开发集、测试集的划分(Train/dev/test distributions)

1.6开发集和测试集的大小 (Size of the dev and test sets)

1.7什么时候改变开发集/测试集和评估指标(When to change dev/test sets and metrics)

1.8为什么是人的表现 (Why human-level performance?)(未看)

1.9可避免偏差(Avoidable bias)(未看)

1.10 理解人类的表现 (Understanding human-level performance)(未看)

1.11 超过人类的表现(Surpassing human-level performance)(未看)

1.12改善你的模型表现 (Improving your model performance)(未看)

2,第二周:机器学习策略(2)(ML Strategy (2))(未看)

四,第四门课 卷积神经网络(Convolutional Neural Networks)

1,第一周 卷积神经网络(Foundations of Convolutional Neural Networks)

1.1计算机视觉(Computer vision)

1.2边缘检测示例(Edge detection example)

1.3更多边缘检测内容(More edge detection)

1.4填充(Padding)

1.5 卷积步长(Strided convolutions)

1.6三维卷积(Convolutions over volumes)

1.7单层卷积网络(One layer of a convolutional network)

1.8简单卷积网络示例(A simple convolution network example)

1.9 池化层(Pooling layers)

1.10卷积神经网络示例(Convolutional neural network example)

1.11为什么使用卷积?(Why convolutions?)

2,第二周 深度卷积网络:实例探究(Deep convolutional models: case studies)

2.1为什么要进行实例探究?(Why look at case studies?)

2.2经典网络(Classic networks)

2.3残差网络(Residual Networks (ResNets))

2.4残差网络为什么有用?(Why ResNets work?)

2.5网络中的网络以及 1×1 卷积(Network in Network and 1×1 convolutions)

2.6谷歌 Inception 网络简介(Inception network motivation)

2.7Inception 网络(Inception network)

2.8使用开源的实现方案(Using open-source implementations)

2.9迁移学习(Transfer Learning)

2.10数据扩充(Data augmentation)

2.11计算机视觉现状(The state of computer vision)

3,第三周 目标检测(Object detection) (未看)

4,第四周 特殊应用:人脸识别和神经风格转换(Special applications: Face recognition &Neural style transfer)(未看)

五,第五门课 序列模型(Sequence Models)(未看)

更多推荐

语义分割入门之路

本文发布于:2024-03-23 14:59:11,感谢您对本站的认可!
本文链接:https://www.elefans.com/category/jswz/34/1739517.html
版权声明:本站内容均来自互联网,仅供演示用,请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系,我们将在24小时内删除。
本文标签:语义   之路   入门

发布评论

评论列表 (有 0 条评论)
草根站长

>www.elefans.com

编程频道|电子爱好者 - 技术资讯及电子产品介绍!