Skip to content

Commit

Permalink
Site updated: 2023-08-09 15:47:28
Browse files Browse the repository at this point in the history
  • Loading branch information
@leafy-lee
leafy-lee committed Aug 9, 2023
1 parent bfd017f commit da05cf0
Show file tree
Hide file tree
Showing 15 changed files with 28 additions and 19 deletions.
2 changes: 1 addition & 1 deletion 2023/07/21/Baseline阅读与修改/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,7 @@
可以发现数据中尺寸总体只有以下几种形式：
{(168, 168, 82, 1), (128, 128, 47, 1), (256, 256, 207, 1), (256, 256, 81, 1), (400, 400, 109, 1), (128, 128, 540, 1), (128, 128, 63, 1), (1.."><meta name="generator" content="Hexo 5.4.2"><link rel="alternate" href="/atom.xml" title="Hexo" type="application/atom+xml">
</head><body class="is-flex is-flex-direction-column"><header class="header-widget is-flex-shrink-0 is-hidden-mobile"><div class="container is-fullhd is-flex is-justify-content-space-between is-align-items-center is-full-height"><section class="is-hidden-mobile is-flex-shrink-0"><h2><a href="/">leafy's blog</a></h2></section><h3 class="is-hidden-mobile is-family-serif is-full-height is-flex is-align-items-center is-flex-shrink-0"><div class="is-full-height" id="postTopic"><p class="is-full-height is-flex-shrink-0 is-flex is-align-items-center is-justify-content-center">Baseline阅读与修改</p><p class="is-full-height is-flex-shrink-0 is-flex is-align-items-center is-justify-content-center">Click back to the top</p></div></h3><aside class="is-flex-shrink-0"><h3 class="is-inline-block"><a href="/">Home</a></h3><h3 class="is-inline-block"><a href="/about">About</a></h3><h3 class="is-inline-block"><a href="/archives">Archives</a></h3></aside></div></header><header class="is-flex header-widget is-flex-shrink-0 is-align-items-center is-justify-content-center is-hidden-tablet"><h3 class="is-inline-block"><a href="/">Home</a></h3><h3 class="is-inline-block"><a href="/about">About</a></h3><h3 class="is-inline-block"><a href="/archives">Archives</a></h3></header><main><main class="container is-max-widescreen content section post-page pt-4 px-4"><div class="columns is-flex-desktop is-justify-content-center is-flex-direction-row-reverse"><div class="column is-3 is-hidden-mobile"><ol class="toc"><li class="toc-item toc-level-2"><a class="toc-link" href="#Baseline%E9%83%A8%E5%88%86"><span class="toc-text">Baseline部分</span></a><ol class="toc-child"><li class="toc-item toc-level-3"><a class="toc-link" href="#%E6%95%B0%E6%8D%AE%E9%9B%86%E5%A4%84%E7%90%86"><span class="toc-text">数据集处理</span></a><ol class="toc-child"><li class="toc-item toc-level-4"><a class="toc-link" href="#%E6%95%B0%E6%8D%AE%E5%88%86%E6%9E%90"><span class="toc-text">数据分析</span></a></li></ol></li><li class="toc-item toc-level-3"><a class="toc-link" href="#%E6%A8%A1%E5%9E%8B%E5%BB%BA%E7%AB%8B"><span class="toc-text">模型建立</span></a><ol class="toc-child"><li class="toc-item toc-level-5"><a class="toc-link" href="#2D-ResNet"><span class="toc-text">2D-ResNet</span></a></li><li class="toc-item toc-level-5"><a class="toc-link" href="#3D-ResNet"><span class="toc-text">3D-ResNet</span></a></li></ol></li></ol></li><li class="toc-item toc-level-3"><a class="toc-link" href="#%E6%A8%A1%E5%9E%8B%E8%AE%AD%E7%BB%83%E4%B8%8E%E9%AA%8C%E8%AF%81"><span class="toc-text">模型训练与验证</span></a></li><li class="toc-item toc-level-3"><a class="toc-link" href="#%E6%A8%A1%E5%9E%8B%E9%A2%84%E6%B5%8B%E4%B8%8E%E6%8F%90%E4%BA%A4"><span class="toc-text">模型预测与提交</span></a></li></ol></li><li class="toc-item toc-level-2"><a class="toc-link" href="#%E8%A2%AB%E6%8F%90%E9%97%AE%E5%88%B0%E7%9A%84%E9%85%8D%E7%BD%AE%E9%97%AE%E9%A2%98"><span class="toc-text">被提问到的配置问题</span></a><ol class="toc-child"><li class="toc-item toc-level-3"><a class="toc-link" href="#Set-ExecutionPolicy"><span class="toc-text">Set-ExecutionPolicy</span></a></li><li class="toc-item toc-level-3"><a class="toc-link" href="#Jupyter-notebook-%E8%BF%90%E8%A1%8C%E6%97%A0%E5%8F%8D%E5%BA%94%E6%88%96-conda-init-%E5%A4%B1%E8%B4%A5"><span class="toc-text">Jupyter notebook 运行无反应或 conda init 失败</span></a></li><li class="toc-item toc-level-3"><a class="toc-link" href="#GBK-%E7%BC%96%E7%A0%81%E9%97%AE%E9%A2%98"><span class="toc-text">GBK 编码问题</span></a></li></ol></li><li class="toc-item toc-level-2"><a class="toc-link" href="#%E4%B8%80%E4%BA%9B%E5%B0%8FTrick"><span class="toc-text">一些小Trick</span></a><ol class="toc-child"><li class="toc-item toc-level-3"><a class="toc-link" href="#Test-Time-Augment%EF%BC%88TTA%EF%BC%89"><span class="toc-text">Test-Time-Augment（TTA）</span></a></li></ol></li><li class="toc-item toc-level-2"><a class="toc-link" href="#%E9%99%84%E5%BD%95%EF%BC%9AA"><span class="toc-text">附录：A</span></a></li></ol></div><div class="column is-9"><header class="my-4"><a href="/tags/Datawhale%20Related"><i class="tag post-item-tag">Datawhale Related</i></a></header><h1 class="mt-0 mb-1 is-family-serif" id="postTitle">Baseline阅读与修改</h1><time class="has-text-grey" datetime="2023-07-21T09:09:59.000Z">2023-07-21</time><article class="mt-2 post-content"><h2 id="Baseline部分"><a href="#Baseline部分" class="headerlink" title="Baseline部分"></a>Baseline部分</h2><h3 id="数据集处理"><a href="#数据集处理" class="headerlink" title="数据集处理"></a>数据集处理</h3><h4 id="数据分析"><a href="#数据分析" class="headerlink" title="数据分析"></a>数据分析</h4><pre><code class="python">sizeSet = set()
</head><body class="is-flex is-flex-direction-column"><header class="header-widget is-flex-shrink-0 is-hidden-mobile"><div class="container is-fullhd is-flex is-justify-content-space-between is-align-items-center is-full-height"><section class="is-hidden-mobile is-flex-shrink-0"><h2><a href="/">leafy's blog</a></h2></section><h3 class="is-hidden-mobile is-family-serif is-full-height is-flex is-align-items-center is-flex-shrink-0"><div class="is-full-height" id="postTopic"><p class="is-full-height is-flex-shrink-0 is-flex is-align-items-center is-justify-content-center">Baseline阅读与修改</p><p class="is-full-height is-flex-shrink-0 is-flex is-align-items-center is-justify-content-center">Click back to the top</p></div></h3><aside class="is-flex-shrink-0"><h3 class="is-inline-block"><a href="/">Home</a></h3><h3 class="is-inline-block"><a href="/about">About</a></h3><h3 class="is-inline-block"><a href="/archives">Archives</a></h3></aside></div></header><header class="is-flex header-widget is-flex-shrink-0 is-align-items-center is-justify-content-center is-hidden-tablet"><h3 class="is-inline-block"><a href="/">Home</a></h3><h3 class="is-inline-block"><a href="/about">About</a></h3><h3 class="is-inline-block"><a href="/archives">Archives</a></h3></header><main><main class="container is-max-widescreen content section post-page pt-4 px-4"><div class="columns is-flex-desktop is-justify-content-center is-flex-direction-row-reverse"><div class="column is-3 is-hidden-mobile"><ol class="toc"><li class="toc-item toc-level-2"><a class="toc-link" href="#Baseline%E9%83%A8%E5%88%86"><span class="toc-text">Baseline部分</span></a><ol class="toc-child"><li class="toc-item toc-level-3"><a class="toc-link" href="#%E6%95%B0%E6%8D%AE%E9%9B%86%E5%A4%84%E7%90%86"><span class="toc-text">数据集处理</span></a><ol class="toc-child"><li class="toc-item toc-level-4"><a class="toc-link" href="#%E6%95%B0%E6%8D%AE%E5%88%86%E6%9E%90"><span class="toc-text">数据分析</span></a></li></ol></li><li class="toc-item toc-level-3"><a class="toc-link" href="#%E6%A8%A1%E5%9E%8B%E5%BB%BA%E7%AB%8B"><span class="toc-text">模型建立</span></a><ol class="toc-child"><li class="toc-item toc-level-5"><a class="toc-link" href="#2D-ResNet"><span class="toc-text">2D-ResNet</span></a></li><li class="toc-item toc-level-5"><a class="toc-link" href="#3D-ResNet"><span class="toc-text">3D-ResNet</span></a></li></ol></li></ol></li><li class="toc-item toc-level-3"><a class="toc-link" href="#%E6%A8%A1%E5%9E%8B%E8%AE%AD%E7%BB%83%E4%B8%8E%E9%AA%8C%E8%AF%81"><span class="toc-text">模型训练与验证</span></a></li><li class="toc-item toc-level-3"><a class="toc-link" href="#%E6%A8%A1%E5%9E%8B%E9%A2%84%E6%B5%8B%E4%B8%8E%E6%8F%90%E4%BA%A4"><span class="toc-text">模型预测与提交</span></a></li></ol></li><li class="toc-item toc-level-2"><a class="toc-link" href="#%E8%A2%AB%E6%8F%90%E9%97%AE%E5%88%B0%E7%9A%84%E9%85%8D%E7%BD%AE%E9%97%AE%E9%A2%98"><span class="toc-text">被提问到的配置问题</span></a><ol class="toc-child"><li class="toc-item toc-level-3"><a class="toc-link" href="#Set-ExecutionPolicy"><span class="toc-text">Set-ExecutionPolicy</span></a></li><li class="toc-item toc-level-3"><a class="toc-link" href="#Jupyter-notebook-%E8%BF%90%E8%A1%8C%E6%97%A0%E5%8F%8D%E5%BA%94%E6%88%96-conda-init-%E5%A4%B1%E8%B4%A5"><span class="toc-text">Jupyter notebook 运行无反应或 conda init 失败</span></a></li><li class="toc-item toc-level-3"><a class="toc-link" href="#GBK-%E7%BC%96%E7%A0%81%E9%97%AE%E9%A2%98"><span class="toc-text">GBK 编码问题</span></a></li></ol></li><li class="toc-item toc-level-2"><a class="toc-link" href="#%E4%B8%80%E4%BA%9B%E5%B0%8FTrick"><span class="toc-text">一些小Trick</span></a><ol class="toc-child"><li class="toc-item toc-level-3"><a class="toc-link" href="#Test-Time-Augment%EF%BC%88TTA%EF%BC%89"><span class="toc-text">Test-Time-Augment（TTA）</span></a></li></ol></li><li class="toc-item toc-level-2"><a class="toc-link" href="#%E9%99%84%E5%BD%95%EF%BC%9AA"><span class="toc-text">附录：A</span></a></li></ol></div><div class="column is-9"><header class="my-4"><a href="/tags/Datawhale-Related"><i class="tag post-item-tag">Datawhale-Related</i></a></header><h1 class="mt-0 mb-1 is-family-serif" id="postTitle">Baseline阅读与修改</h1><time class="has-text-grey" datetime="2023-07-21T09:09:59.000Z">2023-07-21</time><article class="mt-2 post-content"><h2 id="Baseline部分"><a href="#Baseline部分" class="headerlink" title="Baseline部分"></a>Baseline部分</h2><h3 id="数据集处理"><a href="#数据集处理" class="headerlink" title="数据集处理"></a>数据集处理</h3><h4 id="数据分析"><a href="#数据分析" class="headerlink" title="数据分析"></a>数据分析</h4><pre><code class="python">sizeSet = set()
for path in trainPath:
image = nib.load(path)
imgShape = image.dataobj.shape
Expand Down
Binary file added 2023/08/06/行人身份识别/79069068-f7cebf00-7ce8-11ea-948d-d90d2a40fba5-1691567179104.png
View file
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added 2023/08/06/行人身份识别/image-20230806223356737.png
View file
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added 2023/08/06/行人身份识别/image-20230809144608958.png
View file
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
4 changes: 2 additions & 2 deletions 2023/08/06/行人身份识别/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ <h3 id="1-2-赛事任务"><a href="#1-2-赛事任务" class="headerlink" title="
<h3 id="1-3-评价指标"><a href="#1-3-评价指标" class="headerlink" title="1.3 评价指标"></a>1.3 评价指标</h3><p>本模型依据提交的结果文件，采用Macro-F1进行评价，其中Macro-F1的一种定义如下：</p>
<p><img src="/2023/08/06/%E8%A1%8C%E4%BA%BA%E8%BA%AB%E4%BB%BD%E8%AF%86%E5%88%AB/image-20230806212241289.png" alt="image-20230806212241289"></p>
<p>在“Training algorithms for linear text classifiers（ Lewis, David D., et al. “Training algorithms for linear text classifiers.” SIGIR. Vol. 96. 1996.）”中，作者指出，第一种方式对错误的分布不太敏感。，Macro-F1应当是所有类中F1-score的平均值，即第二种方式才是Macro-F1的计算方式，因此我们使用第二种计算方式。</p>
<h2 id="2-任务分析"><a href="#2-任务分析" class="headerlink" title="2 任务分析"></a>2 任务分析</h2><p>​ <img src="/image-20230809144608958.png" alt="image-20230809144608958"></p>
<h2 id="2-任务分析"><a href="#2-任务分析" class="headerlink" title="2 任务分析"></a>2 任务分析</h2><p>​ <img src="/2023/08/06/%E8%A1%8C%E4%BA%BA%E8%BA%AB%E4%BB%BD%E8%AF%86%E5%88%AB/image-20230809144608958.png" alt="image-20230809144608958"></p>
<ol>
<li><p>可以观察到，这里的图像都是上下顶格的，考虑到边缘在CNN卷积的过程中会存在padding的情况，我们可以考试适当放缩图片，为边缘补充zero padding，保证图像效果。</p>
</li>
Expand Down Expand Up @@ -128,7 +128,7 @@ <h3 id="2-模型建立"><a href="#2-模型建立" class="headerlink" title="2.
</code></pre>
<ol start="2">
<li><p>采用学习率的梯度下降，在最小值点提供更小的学习率，防止不收敛。之前同学们提到的0.2-0.5突然震荡回0.8就可能是这种情况。这里我的损失收敛到了0.05-0.12左右。使用了paddle的stepdecay策略，梯度下降学习率。</p>
<p>​ <img src="/image-20230806223356737.png" alt="image-20230806223356737"></p>
<p>​ <img src="/2023/08/06/%E8%A1%8C%E4%BA%BA%E8%BA%AB%E4%BB%BD%E8%AF%86%E5%88%AB/image-20230806223356737.png" alt="image-20230806223356737"></p>
<pre><code class="python">model = XunFeiNet()
model = model
criterion = nn.CrossEntropyLoss(soft_label=True)
Expand Down
2 changes: 1 addition & 1 deletion archives/2023/07/index.html
View file
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
<!DOCTYPE html><html class="appearance-auto" lang="en"><head><meta charset="UTF-8"><title>leafy's blog</title><meta name="description" content="The instinct of a man is to pursue everything that flies from him, and to fly from all that pursue him."><meta name="viewport" content="width=device-width, minimum-scale=1.0, maximum-scale=1.0, user-scalable=no, initial-scale=1"><!-- Google Analytics --><!-- End Google Analytics -->
<!-- Baidu Analytics --><!-- End Baidu Analytics --><link rel="icon" href="/images/favicon.ico"><link rel="stylesheet" href="/style/common/bulma.css"><link rel="stylesheet" href="/style/base.css"><link rel="stylesheet" href="/style/common/helper.css"><script src="/js/common.js"></script><link rel="stylesheet" href="/style/widget-post-list.css"><meta name="generator" content="Hexo 5.4.2"><link rel="alternate" href="/atom.xml" title="Hexo" type="application/atom+xml">
</head><body class="is-flex is-flex-direction-column"><header class="header-widget is-flex-shrink-0 is-hidden-mobile"><div class="container is-fullhd is-flex is-justify-content-space-between is-align-items-center is-full-height"><section class="is-hidden-mobile is-flex-shrink-0"><h2><a href="/">leafy's blog</a></h2></section><h3 class="is-hidden-mobile is-family-serif is-full-height is-flex is-align-items-center is-flex-shrink-0"><span>Archives · 2023</span></h3><aside class="is-flex-shrink-0"><h3 class="is-inline-block"><a href="/">Home</a></h3><h3 class="is-inline-block"><a href="/about">About</a></h3><h3 class="is-inline-block"><a href="/archives">Archives</a></h3></aside></div></header><header class="is-flex header-widget is-flex-shrink-0 is-align-items-center is-justify-content-center is-hidden-tablet"><h3 class="is-inline-block"><a href="/">Home</a></h3><h3 class="is-inline-block"><a href="/about">About</a></h3><h3 class="is-inline-block"><a href="/archives">Archives</a></h3></header><main><article class="post-container is-flex is-justify-content-center section container is-max-widescreen pt-4 px-2"><div class="columns is-variable is-1-tablet is-3-desktop-only is-2-widescreen is-full-width"><section class="column"><article class="post-item-card"><section class="content post-card-content p-4 pb-5"><header><a href="/tags/Datawhale%20Related"><i class="tag post-item-tag">Datawhale Related</i></a></header><h2 class="mt-4 mb-0 is-family-serif"><a href="/2023/07/21/Baseline%E9%98%85%E8%AF%BB%E4%B8%8E%E4%BF%AE%E6%94%B9/">Baseline阅读与修改</a></h2><time class="has-text-grey" datetime="2023-07-21T09:09:59.000Z">2023-07-21</time><p class="is-flex-grow-2 mt-2">Baseline部分数据集处理数据分析sizeSet = set()
</head><body class="is-flex is-flex-direction-column"><header class="header-widget is-flex-shrink-0 is-hidden-mobile"><div class="container is-fullhd is-flex is-justify-content-space-between is-align-items-center is-full-height"><section class="is-hidden-mobile is-flex-shrink-0"><h2><a href="/">leafy's blog</a></h2></section><h3 class="is-hidden-mobile is-family-serif is-full-height is-flex is-align-items-center is-flex-shrink-0"><span>Archives · 2023</span></h3><aside class="is-flex-shrink-0"><h3 class="is-inline-block"><a href="/">Home</a></h3><h3 class="is-inline-block"><a href="/about">About</a></h3><h3 class="is-inline-block"><a href="/archives">Archives</a></h3></aside></div></header><header class="is-flex header-widget is-flex-shrink-0 is-align-items-center is-justify-content-center is-hidden-tablet"><h3 class="is-inline-block"><a href="/">Home</a></h3><h3 class="is-inline-block"><a href="/about">About</a></h3><h3 class="is-inline-block"><a href="/archives">Archives</a></h3></header><main><article class="post-container is-flex is-justify-content-center section container is-max-widescreen pt-4 px-2"><div class="columns is-variable is-1-tablet is-3-desktop-only is-2-widescreen is-full-width"><section class="column"><article class="post-item-card"><section class="content post-card-content p-4 pb-5"><header><a href="/tags/Datawhale-Related"><i class="tag post-item-tag">Datawhale-Related</i></a></header><h2 class="mt-4 mb-0 is-family-serif"><a href="/2023/07/21/Baseline%E9%98%85%E8%AF%BB%E4%B8%8E%E4%BF%AE%E6%94%B9/">Baseline阅读与修改</a></h2><time class="has-text-grey" datetime="2023-07-21T09:09:59.000Z">2023-07-21</time><p class="is-flex-grow-2 mt-2">Baseline部分数据集处理数据分析sizeSet = set()
for path in trainPath:
image = nib.load(path)
imgShape = image.dataobj.shape
Expand Down
Loading

0 comments on commit da05cf0

Please sign in to comment.