<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[qq_32110859的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/qq_32110859</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; qq_32110859]]></copyright><item><title><![CDATA[LeetCode-剑指Offer刷题记录]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/111878699</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/111878699</guid><author>qq_32110859</author><pubDate>Tue, 29 Dec 2020 02:04:30 +0800</pubDate><description><![CDATA[Leetcode刷题之剑指offfer剑指 Offer 04. 二维数组中的查找剑指 Offer 07. 重建二叉树
这个博客记录leetcode-剑指offer中遇到问题的题目
剑指 Offer 04. 二维数组中的查找

在一个 n * m 的二维数组中，每一行都按照从左到右递增的顺序排序，每一列都按照从上到下递增的顺序排序。请完成一个高效的函数，输入这样的一个二维数组和一个整数，判断数组中是否含有该整数。
示例:
现有矩阵 matrix 如下：
[
[1,   4,  7, 11, 15],
[2, ]]></description><category></category></item><item><title><![CDATA[leetcode-刷题记录]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/105219369</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/105219369</guid><author>qq_32110859</author><pubDate>Tue, 31 Mar 2020 14:07:44 +0800</pubDate><description><![CDATA[leetcode-刷题记录初级算法数组从排序数组中删除重复项买卖股票的最佳时机 II旋转数组存在重复只出现一次的数字两个数组的交集 II加一移动零两数之和有效的数独字符串反转字符串整数反转字符串中的第一个唯一字符有效的字母异位词验证回文字符串字符串转换整数 (atoi)实现 strStr()外观数列最长公共前缀链表反转链表合并两个有序链表回文链表环形链表树二叉树的最大深度验证二叉搜索树对称二叉树二...]]></description><category></category></item><item><title><![CDATA[代码debug]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/104953482</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/104953482</guid><author>qq_32110859</author><pubDate>Wed, 18 Mar 2020 21:01:14 +0800</pubDate><description><![CDATA[import logging
logging.basicConfig(level=logging.DEBUG)

The logging showed me my typo in a Python stack trace!!!
]]></description><category></category></item><item><title><![CDATA[opencv读入图像自动旋转的问题]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/104051541</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/104051541</guid><author>qq_32110859</author><pubDate>Mon, 20 Jan 2020 14:45:50 +0800</pubDate><description><![CDATA[最近做一个图像中文字方向判定的需求，4分类任务。

在处理数据的时候发现，在电脑上显示水平的图片，用opencv-imread读进来的时候，会出现旋转，且旋转方向不定。之前也遇到过类似的问题，但当时任务本身对方向不敏感，且预处理本身也会做左右上下的翻转以及小角度的旋转，这样的数据增强操作，而此次方向分类的任务对方向及其敏感。所以不得不解决这个问题。

图片旋转的原因：

图片本身的矩阵就是旋转的，...]]></description><category></category></item><item><title><![CDATA[乘积最大子序列]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/102927296</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/102927296</guid><author>qq_32110859</author><pubDate>Thu, 14 Nov 2019 10:39:58 +0800</pubDate><description><![CDATA[def find_max_m(x):
    max_p = x[0]
    min_n = x[0]
    res = x[0]
    for i in range(1,len(x)):
        max_p = max(max_p*x[i],min_n*x[i],x[i])
        min_n = min(max_p*x[i],min_n*x[i],x[i])
     ...]]></description><category></category></item><item><title><![CDATA[leetcode 15. 三数之和]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/102656491</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/102656491</guid><author>qq_32110859</author><pubDate>Mon, 21 Oct 2019 01:58:19 +0800</pubDate><description><![CDATA[class Solution(object):
    def threeSum(self, nums):
        """
        :type nums: List[int]
        :rtype: List[List[int]]
        """
        nums.sort()
        res = []
        for k in range...]]></description><category></category></item><item><title><![CDATA[tf.py_func采坑记]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/101286739</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/101286739</guid><author>qq_32110859</author><pubDate>Tue, 24 Sep 2019 15:18:28 +0800</pubDate><description><![CDATA[脑子是个好东西，真的希望我也有一个！！！

用tf.py_func时


tf.py_func(
    func,
    inp,
    Tout,
    stateful=True,
    name=None
)

要千万注意Tout: A list or tuple of tensorflow data types or a single tensorflow data type i...]]></description><category></category></item><item><title><![CDATA[attention_ocr源码]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/100514625</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/100514625</guid><author>qq_32110859</author><pubDate>Tue, 03 Sep 2019 15:48:02 +0800</pubDate><description><![CDATA[主要看sequence_layers.py这个脚本中才是实现了attention+decoder的部分，model中只是个架子。
	sequence_layer中也是直接调用了TF的api，如果想深入理解，还是需要看TF源码。先从sequence_layer入手。
AttentionWithAutoregression继承了Attention，Attention继承了SequenceLayerBa...]]></description><category></category></item><item><title><![CDATA[记录eetcode 6.Z 字形变换 两种解法]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/99700552</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/99700552</guid><author>qq_32110859</author><pubDate>Sun, 18 Aug 2019 02:00:04 +0800</pubDate><description><![CDATA[题目本身难度不大，但也耗费了很长时间。是因为受之前题目的影响，不采取最简单的方法，总是想要用一些所谓的便捷方法。

解法1：


class Solution(object):           
    def convert(self, s, numRows):
        """
        :type s: str
        :type numRows: int
     ...]]></description><category></category></item><item><title><![CDATA[记录leetcode 3. 无重复字符的最长子串两种解法]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/99234602</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/99234602</guid><author>qq_32110859</author><pubDate>Sun, 11 Aug 2019 21:49:51 +0800</pubDate><description><![CDATA[1.暴力枚举解法：


class Solution(object):
    def lengthOfLongestSubstring(self, s):
        """
        :type s: str
        :rtype: int
        """
        if len(s) == 0:
            return 0
        eli...]]></description><category></category></item><item><title><![CDATA[记录leetcode两数相加（2）解法]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/98909822</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/98909822</guid><author>qq_32110859</author><pubDate>Fri, 09 Aug 2019 01:58:59 +0800</pubDate><description><![CDATA[# Definition for singly-linked list.
# class ListNode(object):
#     def __init__(self, x):
#         self.val = x
#         self.next = None

class Solution(object):
    def addTwoNumbers(self, l1, ...]]></description><category></category></item><item><title><![CDATA[转pdf文件中的每一页到图片]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/98731782</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/98731782</guid><author>qq_32110859</author><pubDate>Wed, 07 Aug 2019 10:55:32 +0800</pubDate><description><![CDATA[最近遇到，把pdf文件中的每一页转为图片，这样一个需求。折腾了一天。网上大部分教程需要许多安装包，安装十分麻烦，遇到许多坑。终于找到一个比较简单的方法。

https://tecadmin.net/install-imagemagick-on-linux/

如果第一步出现错误：

sudo apt-get update

安装成功后，直接采用 convert xxx.pdf yyy.img 命...]]></description><category></category></item><item><title><![CDATA[链表反转]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/96919007</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/96919007</guid><author>qq_32110859</author><pubDate>Tue, 23 Jul 2019 00:10:47 +0800</pubDate><description><![CDATA[# Definition for singly-linked list.
# class ListNode(object):
#     def __init__(self, x):
#         self.val = x
#         self.next = None

class Solution(object):
    def reverseList(self, hea...]]></description><category></category></item><item><title><![CDATA[ocr中遇到的相关图像处理概念]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/88028919</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/88028919</guid><author>qq_32110859</author><pubDate>Mon, 08 Apr 2019 14:13:28 +0800</pubDate><description><![CDATA[仿射 affine
	平面内的内移，旋转，错切，缩放，变换后仍在一个平面
		三个点
		仿射变换：向量经过线性变换+平移
		
		
		6个未知数，6个方程，1个点列2个方程，需要3个点坐标
	
	投影 projective
	物体与投影面上的像之间的变换，变换前后不在一个平面
		四个点
		
		12个未知数，12个方程，1个点列3个方程，需要4个点坐标
	
	IOU
	
	
	分类中...]]></description><category></category></item><item><title><![CDATA[tf.train.ExponentialMovingAverage]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/86687183</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/86687183</guid><author>qq_32110859</author><pubDate>Tue, 29 Jan 2019 11:35:50 +0800</pubDate><description><![CDATA[https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/train/ExponentialMovingAverage#top_of_page

通过乘以一个指数衰减，保留变量的moving average

训练模型时，保留训练变量的moving average是有用的。平均的参数做验证的效果要优于直接用final trained的...]]></description><category></category></item><item><title><![CDATA[Early Stopping | but when?]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/86673609</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/86673609</guid><author>qq_32110859</author><pubDate>Mon, 28 Jan 2019 16:34:11 +0800</pubDate><description><![CDATA[https://page.mi.fu-berlin.de/prechelt/Biblio/stop_tricks1997.pdf

摘要
	early stop应该是validation-based，但实际中，总是基于an ad-hoc fashion或是training is stopped interactively
		基于临时的策略或是交互的方式停止训练，交互的方式个人理解就是工程师观察l...]]></description><category></category></item><item><title><![CDATA[docker学习]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/86573898</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/86573898</guid><author>qq_32110859</author><pubDate>Tue, 22 Jan 2019 18:39:44 +0800</pubDate><description><![CDATA[依照官网教程安装docker
	https://docs.docker.com/install/linux/docker-ce/centos/#install-from-a-package  手动安装
		过程中显示缺依赖包也是手动下载，然后安装
	
	docker pull
	因为使用的是内网，就先注册，并且没有测试的hello world可以使用
		拉的就是简单的tensorflow镜像
	...]]></description><category></category></item><item><title><![CDATA[经典cnn之resnet]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/86556415</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/86556415</guid><author>qq_32110859</author><pubDate>Sat, 19 Jan 2019 20:26:40 +0800</pubDate><description><![CDATA[https://arxiv.org/pdf/1512.03385.pdf#page=9&amp;amp;zoom=100,0,157

摘要
	 residual 残余的
		We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gai...]]></description><category></category></item><item><title><![CDATA[经典cnn之mobilenet]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/86552188</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/86552188</guid><author>qq_32110859</author><pubDate>Sat, 19 Jan 2019 17:13:33 +0800</pubDate><description><![CDATA[https://arxiv.org/pdf/1704.04861.pdf

摘要
	mobilenet是一个流线型结构，用了depthwise separable convolutions
		用了两个全局超参，在耗时和准确性之间做了trade off，width multiplier and resolution multiplier
	
	prior work
	开发者可以根据resource...]]></description><category></category></item><item><title><![CDATA[tf.image.resize_bilinear vs cv2.resize]]></title><link>https://blog.youkuaiyun.com/qq_32110859/article/details/86139671</link><guid>https://blog.youkuaiyun.com/qq_32110859/article/details/86139671</guid><author>qq_32110859</author><pubDate>Wed, 09 Jan 2019 11:38:45 +0800</pubDate><description><![CDATA[最近做项目，涉及到tf_serving.

之前考虑将image预处理放在serving过程中做，但是遇到大图片时耗时过长的问题。所以将预处理移出pb文件，用java掉opencv

完成。

遇到问题：统一图片java做预处理和tf.image做预处理的预测结果不一致（一百张图片中有1~2张预测结果不一致）。google之后看到相关issue及blog。

https://hackernoon....]]></description><category></category></item></channel></rss>