<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[PartyPartyAnimal的博客]]></title><description><![CDATA[。。]]></description><link>https://blog.csdn.net/PartyPartyAnimal</link><language>zh-cn</language><generator>https://blog.csdn.net/</generator><copyright><![CDATA[Copyright &copy; PartyPartyAnimal]]></copyright><item><title><![CDATA[Perturbative Neural Networks]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/84566740</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/84566740</guid><author>PartyPartyAnimal</author><pubDate>Wed, 28 Nov 2018 16:11:09 +0800</pubDate><description><![CDATA[The perturbation layer does away with convolution in the traditional sense and instead computes its response as a weighted linear combination of non-linearly activated additive noise perturbed inputs....]]></description><category></category></item><item><title><![CDATA[Large Kernel Matters—— Improve Semantic Segmentation by Global Convolutional Network]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/84326735</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/84326735</guid><author>PartyPartyAnimal</author><pubDate>Wed, 21 Nov 2018 17:29:31 +0800</pubDate><description><![CDATA[The large kernel (and effective receptive field) plays an important role when we have to perform the classification and localization tasks simultaneously.（当我们必须同时执行分类和定位任务时，大内核（和有效的感知野）起着重要作用）
we prop...]]></description><category></category></item><item><title><![CDATA[Pyramid Attention Network for Semantic Segmentation]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/84202711</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/84202711</guid><author>PartyPartyAnimal</author><pubDate>Sun, 18 Nov 2018 17:54:11 +0800</pubDate><description><![CDATA[论文地址：https://arxiv.org/pdf/1805.10180.pdf
代码地址（？）：https://github.com/xgmiao/Pyramid-Attention-Networks
摘要：
提出金字塔注意网络（PAN）来利用全局语境信息在语义分割中的影响，结合注意机制和空间金字塔来提取精确的像素标注密集特征，而不是使用复杂的扩散卷积和人工设计的解码器网络。
引入了一个Fea...]]></description><category></category></item><item><title><![CDATA[Deep GrabCut for Object Selection]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/84143674</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/84143674</guid><author>PartyPartyAnimal</author><pubDate>Fri, 16 Nov 2018 18:02:08 +0800</pubDate><description><![CDATA[论文地址：https://arxiv.org/pdf/1707.00243.pdf
代码地址：https://github.com/jfzhang95/DeepGrabCut-PyTorch
（主要就是提出方法的前两步！！！！）
大多数以前基于边界框的分割方法都假设边界框很准，但实际上经常不准。在本文中，提出了一种新的分割方法，它使用矩形框作为软约束，将其转换为欧几里德距离图，将图像与这些距离图c...]]></description><category></category></item><item><title><![CDATA[Memory-Efficient Implementation of DenseNets]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/84108245</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/84108245</guid><author>PartyPartyAnimal</author><pubDate>Thu, 15 Nov 2018 17:44:27 +0800</pubDate><description><![CDATA[**论文地址：**https://arxiv.org/abs/1707.06990
**pytorch实现：**https://github.com/gpleiss/efficient_densenet_pytorch
**tensorflow实现：**https://github.com/joeyearsley/efficient_densenet_tensorflow
属于实现方式的改进，不是...]]></description><category></category></item><item><title><![CDATA[【tensorflow】tf.identity()]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/81348145</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/81348145</guid><author>PartyPartyAnimal</author><pubDate>Wed, 01 Aug 2018 21:01:59 +0800</pubDate><description><![CDATA[常与tf.control_dependencies(self, control_inputs)配合使用，只有当这个上下文管理器中的是一个操作时，control_inputs才会执行。




x = tf.Variable(0.0)
x_plus_1 = tf.assign_add(x, 1)

with tf.control_dependencies([x_plus_1]):
    y = x...]]></description><category></category></item><item><title><![CDATA[【tensorflow】tf.get_variable()和tf.Variable()的区别]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/81347874</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/81347874</guid><author>PartyPartyAnimal</author><pubDate>Wed, 01 Aug 2018 20:44:53 +0800</pubDate><description><![CDATA[1.tf.Variable()



tf.Variable(initial_value, trainable=True, collections=None, validate_shape=True, name=None)
##initial_value为变量的初始值


tf.get_variable




tf.get_variable(name,  shape, initializer):...]]></description><category></category></item><item><title><![CDATA[【tensorflow】命令行参数解析]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/81333026</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/81333026</guid><author>PartyPartyAnimal</author><pubDate>Wed, 01 Aug 2018 16:24:01 +0800</pubDate><description><![CDATA[1. tf.app.flags，用于支持接受命令行传递参数



import tensorflow as tf
#第一个是参数名称，第二个参数是默认值，第三个是参数描述
tf.app.flags.DEFINE_string('str_name', 'def_v_1',&quot;descrip1&quot;)
tf.app.flags.DEFINE_integer('int_name', 10,&quot;descript2...]]></description><category></category></item><item><title><![CDATA[CNN 不变性]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/80991866</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/80991866</guid><author>PartyPartyAnimal</author><pubDate>Tue, 10 Jul 2018 20:48:14 +0800</pubDate><description><![CDATA[CNN的平移不变性：由于全局共享权值和pool操作（明显的特征被保存下来，与位置没有关系？）

CNN的尺度不变性：没有或者说具有一定的不变性（尺度变化不大）

CNN的旋转不变性：基本是没有的，实验证明添加旋转样本是一种可靠的样本增强策略，能增强模型对旋转的鲁棒性（分类中很适合，多种角度都对应的是同一个目标；对于语义分割来说，x和y需要同时进行旋转，只是为了增加样本。）...]]></description><category></category></item><item><title><![CDATA[目标检测]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79980892</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79980892</guid><author>PartyPartyAnimal</author><pubDate>Tue, 17 Apr 2018 21:10:12 +0800</pubDate><description><![CDATA[目标检测二十年 
http://www.tuan18.org/thread-509-1-1.html]]></description><category></category></item><item><title><![CDATA[Python图像处理库PIL]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79980868</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79980868</guid><author>PartyPartyAnimal</author><pubDate>Tue, 17 Apr 2018 21:08:58 +0800</pubDate><description><![CDATA[Python图像处理库PIL的基本概念介绍  
https://blog.csdn.net/icamera0/article/details/50647465 
Python图像处理库PIL的Image模块介绍（一） 
https://blog.csdn.net/icamera0/article/details/50654910 
Python图像处理库PIL的Image模块介绍（二） 
http...]]></description><category></category></item><item><title><![CDATA[25.  复杂链表的复制]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79719456</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79719456</guid><author>PartyPartyAnimal</author><pubDate>Tue, 27 Mar 2018 21:15:35 +0800</pubDate><description><![CDATA[主程序包含三大步： 
（1）将链表的每个结点复制下来，并将依次连接在对应结点的后面，将其random域设为空；（2）拷贝random域：复制结点的random域是原始结点对应的random结点的下一个；（3）将链表拆分为两部分：遍历链表中的每个元素，将每个元素的指针指向下下个元素。（这里要定义返回的复制链表的头）/*
struct RandomListNode {
    int label;]]></description><category></category></item><item><title><![CDATA[19. 顺时针打印矩阵]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79701766</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79701766</guid><author>PartyPartyAnimal</author><pubDate>Mon, 26 Mar 2018 18:09:10 +0800</pubDate><description><![CDATA[class Solution {
public:
    vector<int> printMatrix(vector<vector<int> > matrix) {
        int row = matrix.size();
        int col = matrix[0].size();
        vector<int> res;
        // 首先判断矩阵是否合法]]></description><category></category></item><item><title><![CDATA[17. 树的子结构]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79701714</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79701714</guid><author>PartyPartyAnimal</author><pubDate>Mon, 26 Mar 2018 18:04:33 +0800</pubDate><description><![CDATA[/*
struct TreeNode {
    int val;
    struct TreeNode *left;
    struct TreeNode *right;
    TreeNode(int x) :
            val(x), left(NULL), right(NULL) {
    }
};*/
class Solution {
public:
    bool]]></description><category></category></item><item><title><![CDATA[15]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79658846</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79658846</guid><author>PartyPartyAnimal</author><pubDate>Thu, 22 Mar 2018 19:48:14 +0800</pubDate><description><![CDATA[首先判断链表是否为空（没写会出现段错误）

定义三个指针，cur指的是当前元素，pre指的是前一个元素，nextcur需要保存的是下一个元素，也就是下一次循环时的当前元素。



/*
struct ListNode {
    int val;
    struct ListNode *next;
    ListNode(int x) :
            val(x), next(NU...]]></description><category></category></item><item><title><![CDATA[13  调整数组顺序使奇数位于偶数前面]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79654433</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79654433</guid><author>PartyPartyAnimal</author><pubDate>Thu, 22 Mar 2018 15:07:46 +0800</pubDate><description><![CDATA[需要记录排好序的最后一个奇数的位置odd_end 
然后遇到奇数就一个一个向前交换，将该奇数交换到odd_end+1位置，更新odd_end=odd_end+1

class Solution {
public:
    void reOrderArray(vector&amp;lt;int&amp;gt; &amp;amp;array) {
        int array_len = array.size();
 ...]]></description><category></category></item><item><title><![CDATA[14  链表中倒数第k个结点]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79654370</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79654370</guid><author>PartyPartyAnimal</author><pubDate>Thu, 22 Mar 2018 15:03:47 +0800</pubDate><description><![CDATA[定义两个指针start和end，让start先走k步（在走的过程中，需要判断start是否指向NULL，指向的话说明链表长度小于k，直接返回NULL）,然后再判断当前start是否指向NULL，指向的话说明量表长度等于k，直接返回end，否则令它们同时往后移动，当start到达链表末尾时，end所指即为链表中倒数第k个结点。

/*
struct ListNode {
    int val;
 ...]]></description><category></category></item><item><title><![CDATA[6  旋转数组的最小数字]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79633434</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79633434</guid><author>PartyPartyAnimal</author><pubDate>Tue, 20 Mar 2018 22:57:45 +0800</pubDate><description><![CDATA[class Solution {
public:
    int minNumberInRotateArray(vector&amp;lt;int&amp;gt; rotateArray) {
        int len = rotateArray.size();
        if(len == 0) return 0;
        int left = 0, right = len - 1;
   ...]]></description><category></category></item><item><title><![CDATA[5  用两个栈实现队列]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79628721</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79628721</guid><author>PartyPartyAnimal</author><pubDate>Tue, 20 Mar 2018 17:17:43 +0800</pubDate><description><![CDATA[class Solution
{
public:
    void push(int node) {
        stack1.push(node);        
    }

    int pop() {
        if(stack2.empty()){
            if(stack1.empty()) return 0;
            while(!sta...]]></description><category></category></item><item><title><![CDATA[4  重建二叉树]]></title><link>https://blog.csdn.net/PartyPartyAnimal/article/details/79628227</link><guid>https://blog.csdn.net/PartyPartyAnimal/article/details/79628227</guid><author>PartyPartyAnimal</author><pubDate>Tue, 20 Mar 2018 16:54:52 +0800</pubDate><description><![CDATA[/**
 * Definition for binary tree
 * struct TreeNode {
 *     int val;
 *     TreeNode *left;
 *     TreeNode *right;
 *     TreeNode(int x) : val(x), left(NULL), right(NULL) {}
 * };
 */
class Soluti...]]></description><category></category></item></channel></rss>