<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[嘤嘤怪赚钱养妈妈]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/jiangyutongyangyi</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; jiangyutongyangyi]]></copyright><item><title><![CDATA[Multi-View Saliency Guided Deep Neural Network for 3-D Object Retrieval and Classification]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/107108808</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/107108808</guid><author>jiangyutongyangyi</author><pubDate>Sat, 04 Jul 2020 00:02:15 +0800</pubDate><description><![CDATA[&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;Multi-View Saliency Guided Deep Neural Network for 3-D Object Retrieval and Classification 是2020年发表在IEEE  transaction on multimedia 上的论文。
这篇论文的目的是解决以下两个问题

如何充分利用多视角语境信息，同时实现代表性视图的选择和相似性测度
如何在没有特定摄像机设置的情况下发现多视图序列的判]]></description><category></category></item><item><title><![CDATA[RuntimeError: Expected object of scalar type Float but got scalar type Double for argument #4 'mat1']]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105539186</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105539186</guid><author>jiangyutongyangyi</author><pubDate>Wed, 15 Apr 2020 17:39:26 +0800</pubDate><description><![CDATA[&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;1在学习《动手学深度学习》(Dive-into-DL-PyTorch)的时候。出现了改错误。
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;RuntimeError: Expected object of scalar type Float but got scalar typ...]]></description><category></category></item><item><title><![CDATA[pytorch index_select用法]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105516543</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105516543</guid><author>jiangyutongyangyi</author><pubDate>Tue, 14 Apr 2020 17:44:33 +0800</pubDate><description><![CDATA[&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;在看动手学深度学习 pytorch版本的时候，看到其中使用了index_select方法。感觉这是有关较为常用和有用的方法。所以需要弄懂。
用法如下：
torch.index_select(input, dim, index, out=None) → Tensor

参数input(Tensor)：表示...]]></description><category></category></item><item><title><![CDATA[juypter notebook中调用其他模块]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105498262</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105498262</guid><author>jiangyutongyangyi</author><pubDate>Mon, 13 Apr 2020 22:08:40 +0800</pubDate><description><![CDATA[&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;jupter notebook很方便，但是它的后缀是.ipynb，使用不能直接被python解析器解析。官网特供了一个简单的方法就是，创建一个解析包，只要先导入了该解析包，就可以在jupter notebook中调用自己写的模块。
步骤如下：
1 建立Ipynb_importer.py文件
import...]]></description><category></category></item><item><title><![CDATA[np.random一系列(np.random.normal()、np.random.randint、np.random.randn、np.random.rand)]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105494074</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105494074</guid><author>jiangyutongyangyi</author><pubDate>Mon, 13 Apr 2020 20:18:03 +0800</pubDate><description><![CDATA[&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;在使用numpy的时候，我们经常会使用到np.random一系列的有关函数,来创建ndarray 数组。random代表随机的意思，指ndarray中的数是随机数。后面的函数表示随机生成的ndarray需要符合什么样的条件。因为其太多，所以容易弄混淆下面将其常用的几个列出来（后续遇到新的不断...]]></description><category></category></item><item><title><![CDATA[jupyter notebook自动补全]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105460970</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105460970</guid><author>jiangyutongyangyi</author><pubDate>Sat, 11 Apr 2020 22:29:16 +0800</pubDate><description><![CDATA[&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;jupyter很好用，但是有时候希望像pycharm这些idea一样可以自动补全。只需要安装上插件它也是可以的。
如下步骤:
1. 执行以下两命令，我是在anaconda中执行的
pip install jupyter_contrib_nbextensions 
jupyter ...]]></description><category></category></item><item><title><![CDATA[将网页另存为PDF(如csdn的博客)]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105460281</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105460281</guid><author>jiangyutongyangyi</author><pubDate>Sat, 11 Apr 2020 21:44:06 +0800</pubDate><description><![CDATA[如下步骤：
1. 先进入网页开发者模式，一般按F12，网页显示如下：

2. 点击console:

3. 然后输入：
(function(){
$("#side").remove();
$("#comment_title, #comment_list, #comment_bar, #comment_form, .announce, #ad_cen, #ad_bot").remove();
$("...]]></description><category></category></item><item><title><![CDATA[高斯滤波/高斯模糊(Gaussian blur)和高斯噪声(Gaussian noise)]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105281262</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105281262</guid><author>jiangyutongyangyi</author><pubDate>Sat, 04 Apr 2020 21:15:50 +0800</pubDate><description><![CDATA[&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;在看CV方面的论文的时候很多时候都会出现高斯滤波/高斯模糊(Gaussian blur)。所以打算把高斯滤波弄明白。
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;高斯模糊有叫做高斯平滑也叫做高斯滤波，是不是好多名字。不过这也正常，比如说我们人可能有很多...]]></description><category></category></item><item><title><![CDATA[Feedback Network for Image Super-Resolution]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105257409</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/105257409</guid><author>jiangyutongyangyi</author><pubDate>Thu, 02 Apr 2020 21:56:11 +0800</pubDate><description><![CDATA[Feedback Network for Image Super-Resolution 是CVPR2019的论文
论文地址：https://arxiv.org/pdf/1903.09814.pdf
pytorch代码地址：https://github.com/Paper99/SRFBN_CVPR19
这篇论文做出的贡献

在图像超分辨率上提出了应该网络super-resolution feedba...]]></description><category></category></item><item><title><![CDATA[machine learing in action(机器学习实战)中 一些代码问题]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104742021</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104742021</guid><author>jiangyutongyangyi</author><pubDate>Sun, 08 Mar 2020 22:21:21 +0800</pubDate><description><![CDATA[因为这本书是用python2写的有些地方和python有一点不一样，因而报错的地方或是有些小问题，记录下来。一直更新。
1. UnicodeDecodeError: ‘gbk’ codec can’t decode byte 0x80 in position 205: illegal multibyte sequence
不能解码gbk的内容，修改如下：
file= open(‘order.log...]]></description><category></category></item><item><title><![CDATA[论文 Meta-SR: A Magnification-Arbitrary Network for Super-Resolution阅读笔记]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104504705</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104504705</guid><author>jiangyutongyangyi</author><pubDate>Tue, 25 Feb 2020 23:46:43 +0800</pubDate><description><![CDATA[Meta-SR: A Magnifification-Arbitrary Network for Super-Resolution
是CVPR2019的论文，由中科院、CASIA、旷视、清华等联合发布表。
论文地址：http://arxiv.org/abs/1903.00875v1
pytorch代码地址：https://github.com/XuecaiHu/Meta-SR-Pytorch
一 ...]]></description><category></category></item><item><title><![CDATA[插值（五）Bicubic interpolation（双三次插值)]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104482467</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104482467</guid><author>jiangyutongyangyi</author><pubDate>Mon, 24 Feb 2020 21:05:32 +0800</pubDate><description><![CDATA[感觉了解插值就像是打怪一样，现在终于打到这个终极大boss了。插值最后一篇。

  上一篇我们讲了cubic interpolation（三次插值），这一篇我们来讲bicubic interpolation(双线性插值)

  双三次插值：是二维空间中最常用的插值方法。在这种方法中，函数f在点 (x,y) 的值可以通过矩形网格中最近的十六个采样点的加权平均得到，在这里需要使用两...]]></description><category></category></item><item><title><![CDATA[插值(三) nearest-neighbor(最近邻插值)]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104482342</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104482342</guid><author>jiangyutongyangyi</author><pubDate>Mon, 24 Feb 2020 18:16:34 +0800</pubDate><description><![CDATA[这个方法是很简单的，但是用于图像插值的时候，放大后的图片会比较模糊。也用于实时3D图像的重建和选择纹理表面的颜色值

 这个不需要复杂的计算。



当在A区域时选择(i,j)处的值，当在B区域时选择(i+,j)处的值，当在C区域时选择(i,j+1)处的值，当在D区域时选择(i+1,j+1)处的值。就是选择离在比较近的点的值。
...]]></description><category></category></item><item><title><![CDATA[插值（四）cubic interpolation(三次插值)]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104470184</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104470184</guid><author>jiangyutongyangyi</author><pubDate>Sun, 23 Feb 2020 23:49:33 +0800</pubDate><description><![CDATA[nearest-neighbor interpolation比较简单，主要应用于实时3D渲染和选择纹理表面颜色值。

不需要计算。

选择离自己最近的像素点，缺点就是放大后的图片不是很清晰。



落在A区域的选择（i,j）,落在B区域选择（i+1,j）点。下一篇讲cubic interpolation（三次插值）


...]]></description><category></category></item><item><title><![CDATA[插值（二）Bilinear interpolation(双线性插值)]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104469715</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104469715</guid><author>jiangyutongyangyi</author><pubDate>Sun, 23 Feb 2020 23:30:01 +0800</pubDate><description><![CDATA[上次说到linear interpolation（这可能是一部连续剧吧，哈哈）。

  bilinear interpolation 是先在一个方向用linear interpolation。再在另外一个方向用linear interpolation。虽然每一步在采样值和方向上都是线性的。但是整体不是线性的。

   看算法：

    如果我们不知道 functio...]]></description><category></category></item><item><title><![CDATA[插值（一）Linear interpolation(线性插值)]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104469039</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104469039</guid><author>jiangyutongyangyi</author><pubDate>Sun, 23 Feb 2020 22:54:47 +0800</pubDate><description><![CDATA[看超分辨率的论文，看到Bicubic Interpolation（双三次插值），去看文档，说是基于cubic Interpolationnterpolation(三次插值)。

然后文档里面写The interpolated surface is smoother than corresponding surfaces obtained by bilinear interpolation o...]]></description><category></category></item><item><title><![CDATA[Sub-pixel Convolution]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104467093</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104467093</guid><author>jiangyutongyangyi</author><pubDate>Sun, 23 Feb 2020 21:03:13 +0800</pubDate><description><![CDATA[Sub-pixel Convolution叫做子像素卷积，一般用于SR(超分辨率)中。用于在网络的末端对特征图进行放大。常见的方法有直接上采样，双线性插值等。

 Sub-pixel Convolution的做法如下。



  前面是lower-resolution image进行普通卷积，最后一步彩色的地方就是，Sub-pixel Convolution。如果把一张图片放大一张...]]></description><category></category></item><item><title><![CDATA[python 打开csv文件报错UnicodeDecodeError: 'utf-8' codec can't decode byte 0xc8 in position 0: invalid cont]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104451482</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104451482</guid><author>jiangyutongyangyi</author><pubDate>Sat, 22 Feb 2020 22:30:21 +0800</pubDate><description><![CDATA[UnicodeDecodeError: 'utf-8' codec can't decode byte 0xc8 in position 0: invalid continuation byte


这个错误就是utf-8”编解码器无法对位置0中的字节0xc8进行解码

我看网上说把csv文件编码格式改了，最后失败

解决方法是data = pd.read_csv('mydata4.csv',e...]]></description><category></category></item><item><title><![CDATA[pyhton使用正则表达式时出现错误 'NoneType' object has no attribute 'group']]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104445796</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104445796</guid><author>jiangyutongyangyi</author><pubDate>Sat, 22 Feb 2020 16:28:53 +0800</pubDate><description><![CDATA[错误的中文意思是空类型没有属性group

当时我写的正则表达式为

match = re.search(r"(\d+), Date: (.+)","ID: 0012819,Date: Feb/12/2017")
print(match)
print(match.group())
print(match.group(1))
print(match.group(2))

看出什么错了吗，其实第一...]]></description><category></category></item><item><title><![CDATA[python三引号（“”“ ”“”）]]></title><link>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104444997</link><guid>https://blog.youkuaiyun.com/jiangyutongyangyi/article/details/104444997</guid><author>jiangyutongyangyi</author><pubDate>Sat, 22 Feb 2020 15:58:44 +0800</pubDate><description><![CDATA["""" """一表示个字符串跨多行，字符串中可以包含换行符、制表符以及其他特殊字符

string = """dog runs to me.
I run to dog
"""

输出

如果是

string = """dog runs to me.
I run to dog
"""

I 前面有一个打了空格那么也会有空格的


...]]></description><category></category></item></channel></rss>