<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[Pleadge的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/Pleadge</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; Pleadge]]></copyright><item><title><![CDATA[如何爆优快云的菊花]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/132576344</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/132576344</guid><author>Pleadge</author><pubDate>Wed, 30 Aug 2023 09:53:01 +0800</pubDate><description><![CDATA[再请一天假啊,SB]]></description><category></category></item><item><title><![CDATA[RuntimeError: received 0 items of ancdata]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/131120330</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/131120330</guid><author>Pleadge</author><pubDate>Fri, 09 Jun 2023 09:19:30 +0800</pubDate><description><![CDATA[为了巴尔!]]></description><category></category></item><item><title><![CDATA[计算机视觉Baseline网络串烧（AlexNet、NIN、VGG、GoogLeNet、Inception、ResNet、InceptionResNet、DenseNet、SENet、ResNeXt）]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/126723564</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/126723564</guid><author>Pleadge</author><pubDate>Fri, 09 Sep 2022 16:58:07 +0800</pubDate><description><![CDATA[计算机视觉基础网络核心总结]]></description><category></category></item><item><title><![CDATA[一文读懂CRNN（基于图像序列识别的端到端可训练网络）]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/126417998</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/126417998</guid><author>Pleadge</author><pubDate>Fri, 19 Aug 2022 10:48:20 +0800</pubDate><description><![CDATA[CRNN网络介绍]]></description><category></category></item><item><title><![CDATA[论文翻译VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITION]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/126345875</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/126345875</guid><author>Pleadge</author><pubDate>Wed, 17 Aug 2022 09:30:55 +0800</pubDate><description><![CDATA[优快云，我爱你家人]]></description><category></category></item><item><title><![CDATA[论文翻译ImageNet Classification with Deep Convolutional Neural Networks]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/126143407</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/126143407</guid><author>Pleadge</author><pubDate>Mon, 15 Aug 2022 11:14:18 +0800</pubDate><description><![CDATA[优快云， 我爱你家人]]></description><category></category></item><item><title><![CDATA[论文翻译(二) CRNN 基于图像序列的识别的识别的端到端可训练神经网络及其在场景文字识别的应用]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/126170738</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/126170738</guid><author>Pleadge</author><pubDate>Tue, 09 Aug 2022 15:16:32 +0800</pubDate><description><![CDATA[优快云，我爱你家人]]></description><category></category></item><item><title><![CDATA[论文翻译（一） GoogleNet-V1 ：Going Depper with convolutions]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/125861744</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/125861744</guid><author>Pleadge</author><pubDate>Sat, 30 Jul 2022 17:09:42 +0800</pubDate><description><![CDATA[优快云，我爱你家人]]></description><category></category></item><item><title><![CDATA[ubuntu 开机自动挂载硬盘]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/124985755</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/124985755</guid><author>Pleadge</author><pubDate>Thu, 26 May 2022 15:42:49 +0800</pubDate><description><![CDATA[首先运行
sudo gedit /etc/fstab
之后用软件disk或者运行命令fdisk -l
[设备ID] [默认挂载点] [文件系统类型 ] [挂载选项 ] [开机检查] [系统:0 交换:1 存储:2]
示例:
# M2_DISK_Z
UUID=1E9EC7C79EC79625 /media/liuhong/M2_DISK_Z ntfs defaults 0 2
# SATA_DISK_W
UUID=4498463A98462B2E /media/liuhong/SATA_DISK_W ntfs]]></description><category></category></item><item><title><![CDATA[更换python链接并升级对应pip]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/124611483</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/124611483</guid><author>Pleadge</author><pubDate>Fri, 06 May 2022 15:35:09 +0800</pubDate><description><![CDATA[1.首先将python3软链接删除
rm -rf /usr/bin/python3
2.之后链接python3.7到python上面
ln -s /usr/bin/python3.7 /usr/bin/python3
3.安装对应pip
sudo apt install python3-pip
或
python3 get-pip.py
或
python3 -m pip install --upgrade pip	
(亲测最后一个好用)

...]]></description><category></category></item><item><title><![CDATA[论文阅读瞎记(五) CenterNet: Objects as Points 2019]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/123654767</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/123654767</guid><author>Pleadge</author><pubDate>Wed, 30 Mar 2022 15:07:09 +0800</pubDate><description><![CDATA[概要
在图片中检测识别物体作为轴对齐（axis-aligned）的bbox。大多数成功的检测器枚举了临近的潜在物体位置列表并且对其进行分类。这是浪费，不高效并且需要额外的后处理的。这篇文章中作者使用了不同的方法。作者将每一个物体建模为一个点（也就是物体的中心点）作者的检测器使用了关键点估计来找到中心点并且回归到其他物体属性中去，比如尺寸，3D位置，方向甚至动作。作者基于中心点的方法，CenterNet是端到端可微分的，简单快速且相比anchor base方法更准确。CenterNet在COCO数据集中达到了]]></description><category></category></item><item><title><![CDATA[论文阅读瞎记(四) Cascade R-CNN: Delving into High Quality Object Detection 2017]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/123541472</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/123541472</guid><author>Pleadge</author><pubDate>Tue, 22 Mar 2022 00:17:40 +0800</pubDate><description><![CDATA[概述
在物体检测中1，IOU阈值被用于判定正负样本。在低IOU阈值比如0.5的状态下训练模型经常产生噪音预测，然而检测效果会随着IOU增加而降低。两个主要因素：1.训练时的过拟合，正样本指数消失 2.inference-time mismatch between the
IoUs for which the detector is optimal and those of the input hypotheses.（后处理的问题？）一个单阶段的物体检测器CascadeR-CNN被提出用于解决这些问题。网络]]></description><category></category></item><item><title><![CDATA[论文阅读瞎记(三) EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks 2020]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/123358112</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/123358112</guid><author>Pleadge</author><pubDate>Wed, 16 Mar 2022 17:12:46 +0800</pubDate><description><![CDATA[不清楚的地方：
GPipe网络结构是啥？
CIFAR-100数据集，具体是？
Flowers 是花分类数据集?
摘要
卷积神经网络开发时大多资源预算固定，之后如果有更多的资源就会扩大网络提升效果。这篇论文系统讨论了模型体量变化并发现仔细平衡网络的深度宽度以及分辨率可得到更佳表现。提出了新的网络尺度变化方法，通过简单但高效的compound coefficient.（复合系数）联合变换深度宽度分辨率。并在MobileNets和ResNet上论证有效。
更进一步使用了神经网络结构搜索来设计新的基线网络，并且变换]]></description><category></category></item><item><title><![CDATA[论文阅读瞎记(二)EfficientDet: Scalable and Efficient Object Detection 2020]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/123307351</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/123307351</guid><author>Pleadge</author><pubDate>Tue, 08 Mar 2022 16:36:20 +0800</pubDate><description><![CDATA[论文阅读瞎记(二)EfficientDet: Scalable and Efficient Object Detection 2020
翻译加自己概括，有问题的地方请指正
不清楚：
PANet
NAS-FPN
计算机视觉中模型的效率越发重要，本文系统地学习基于检测的神经网络架构选择，并提出几个关键优化来提升效率。首先是提出了一种加权双向FPN（BiFPN），它能够快速简单地进行多尺度特征融合。第二，我们提出了一种复合标度方法，该方法可以同时对所有主干网络、特征网络和盒类预测网络的分辨率、深度和宽度进行统一标]]></description><category></category></item><item><title><![CDATA[论文阅读瞎记(一)FCOS: Fully Convolutional One-Stage Object Detection (CVPR2019)]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/123207963</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/123207963</guid><author>Pleadge</author><pubDate>Thu, 03 Mar 2022 15:31:21 +0800</pubDate><description><![CDATA[swin主干网络
asff
CBAM
SPPF
FCN 全卷积网络，语义分割，深度重建，关键点检测   Fully convolutional networks for semantic segmentation.
CornerNet
训练时不需要计算重叠，后处理只有NMS，也不需要与bbox相关的超参数
基于anchor base方法的缺点：
1.效果对于尺寸，长宽比以及box数量敏感。依赖相关超参数
2.适用场景固定，对于不同场景尤其是小尺寸物体适应性差，需要重新调节超参数
3.保证高recall需要输]]></description><category></category></item><item><title><![CDATA[cmake安装]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/117174644</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/117174644</guid><author>Pleadge</author><pubDate>Sun, 23 May 2021 00:51:10 +0800</pubDate><description><![CDATA[os：ubuntu 12.04
cmake版本 2.8.10.2
1、下载 http://www.cmake.org/cmake/resources/software.html
cmake-2.8.10.2.tar.gz
2、解压
3、安装 三条命令
./bootstrap
make
sudo make install
4、检查是否安装成功：
输入shell命令: cmake --version
给出一下信息表示安装成功：cmake version 2.8.10.2
cmake 会默认安装在 /usr/lo]]></description><category></category></item><item><title><![CDATA[如果你使用 Ubuntu or Debian, 安装 libgtk2.0-dev 以及 pkg-config]]></title><link>https://blog.youkuaiyun.com/Pleadge/article/details/107583678</link><guid>https://blog.youkuaiyun.com/Pleadge/article/details/107583678</guid><author>Pleadge</author><pubDate>Sat, 25 Jul 2020 19:38:37 +0800</pubDate><description><![CDATA[这里写自定义If you are on Ubuntu or Debian, install libgtk2.0-dev and pkg-config目录标题欢迎使用Markdown编辑器新的改变功能快捷键合理的创建标题，有助于目录的生成如何改变文本的样式插入链接与图片如何插入一段漂亮的代码片生成一个适合你的列表创建一个表格设定内容居中、居左、居右SmartyPants创建一个自定义列表如何创建一个注脚注释也是必不可少的KaTeX数学公式新的甘特图功能，丰富你的文章UML 图表FLowchart流程图导出与导]]></description><category></category></item></channel></rss>