589 lines
71 KiB
HTML
589 lines
71 KiB
HTML
|
|
|
|
<!DOCTYPE html>
|
|
<!--[if IE 8]><html class="no-js lt-ie9" lang="en" > <![endif]-->
|
|
<!--[if gt IE 8]><!--> <html class="no-js" lang="en" > <!--<![endif]-->
|
|
<head>
|
|
<meta charset="utf-8">
|
|
|
|
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
|
|
|
<title>torchreid.models.hacnn — torchreid 0.8.1 documentation</title>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<script type="text/javascript" src="../../../_static/js/modernizr.min.js"></script>
|
|
|
|
|
|
<script type="text/javascript" id="documentation_options" data-url_root="../../../" src="../../../_static/documentation_options.js"></script>
|
|
<script type="text/javascript" src="../../../_static/jquery.js"></script>
|
|
<script type="text/javascript" src="../../../_static/underscore.js"></script>
|
|
<script type="text/javascript" src="../../../_static/doctools.js"></script>
|
|
<script type="text/javascript" src="../../../_static/language_data.js"></script>
|
|
<script async="async" type="text/javascript" src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.5/latest.js?config=TeX-AMS-MML_HTMLorMML"></script>
|
|
|
|
<script type="text/javascript" src="../../../_static/js/theme.js"></script>
|
|
|
|
|
|
|
|
|
|
<link rel="stylesheet" href="../../../_static/css/theme.css" type="text/css" />
|
|
<link rel="stylesheet" href="../../../_static/pygments.css" type="text/css" />
|
|
<link rel="index" title="Index" href="../../../genindex.html" />
|
|
<link rel="search" title="Search" href="../../../search.html" />
|
|
</head>
|
|
|
|
<body class="wy-body-for-nav">
|
|
|
|
|
|
<div class="wy-grid-for-nav">
|
|
|
|
<nav data-toggle="wy-nav-shift" class="wy-nav-side">
|
|
<div class="wy-side-scroll">
|
|
<div class="wy-side-nav-search" >
|
|
|
|
|
|
|
|
<a href="../../../index.html" class="icon icon-home"> torchreid
|
|
|
|
|
|
|
|
</a>
|
|
|
|
|
|
|
|
|
|
<div class="version">
|
|
0.8.1
|
|
</div>
|
|
|
|
|
|
|
|
|
|
<div role="search">
|
|
<form id="rtd-search-form" class="wy-form" action="../../../search.html" method="get">
|
|
<input type="text" name="q" placeholder="Search docs" />
|
|
<input type="hidden" name="check_keywords" value="yes" />
|
|
<input type="hidden" name="area" value="default" />
|
|
</form>
|
|
</div>
|
|
|
|
|
|
</div>
|
|
|
|
<div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="main navigation">
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<ul>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../user_guide.html">How-to</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../datasets.html">Datasets</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../evaluation.html">Evaluation</a></li>
|
|
</ul>
|
|
<p class="caption"><span class="caption-text">Package Reference</span></p>
|
|
<ul>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../pkg/data.html">torchreid.data</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../pkg/engine.html">torchreid.engine</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../pkg/losses.html">torchreid.losses</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../pkg/metrics.html">torchreid.metrics</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../pkg/models.html">torchreid.models</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../pkg/optim.html">torchreid.optim</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../pkg/utils.html">torchreid.utils</a></li>
|
|
</ul>
|
|
<p class="caption"><span class="caption-text">Resources</span></p>
|
|
<ul>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../AWESOME_REID.html">Awesome-ReID</a></li>
|
|
<li class="toctree-l1"><a class="reference internal" href="../../../MODEL_ZOO.html">Model Zoo</a></li>
|
|
</ul>
|
|
|
|
|
|
|
|
</div>
|
|
</div>
|
|
</nav>
|
|
|
|
<section data-toggle="wy-nav-shift" class="wy-nav-content-wrap">
|
|
|
|
|
|
<nav class="wy-nav-top" aria-label="top navigation">
|
|
|
|
<i data-toggle="wy-nav-top" class="fa fa-bars"></i>
|
|
<a href="../../../index.html">torchreid</a>
|
|
|
|
</nav>
|
|
|
|
|
|
<div class="wy-nav-content">
|
|
|
|
<div class="rst-content">
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<div role="navigation" aria-label="breadcrumbs navigation">
|
|
|
|
<ul class="wy-breadcrumbs">
|
|
|
|
<li><a href="../../../index.html">Docs</a> »</li>
|
|
|
|
<li><a href="../../index.html">Module code</a> »</li>
|
|
|
|
<li>torchreid.models.hacnn</li>
|
|
|
|
|
|
<li class="wy-breadcrumbs-aside">
|
|
|
|
</li>
|
|
|
|
</ul>
|
|
|
|
|
|
<hr/>
|
|
</div>
|
|
<div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
|
|
<div itemprop="articleBody">
|
|
|
|
<h1>Source code for torchreid.models.hacnn</h1><div class="highlight"><pre>
|
|
<span></span><span class="kn">from</span> <span class="nn">__future__</span> <span class="k">import</span> <span class="n">absolute_import</span>
|
|
<span class="kn">from</span> <span class="nn">__future__</span> <span class="k">import</span> <span class="n">division</span>
|
|
|
|
<span class="n">__all__</span> <span class="o">=</span> <span class="p">[</span><span class="s1">'HACNN'</span><span class="p">]</span>
|
|
|
|
<span class="kn">import</span> <span class="nn">torch</span>
|
|
<span class="kn">from</span> <span class="nn">torch</span> <span class="k">import</span> <span class="n">nn</span>
|
|
<span class="kn">from</span> <span class="nn">torch.nn</span> <span class="k">import</span> <span class="n">functional</span> <span class="k">as</span> <span class="n">F</span>
|
|
<span class="kn">import</span> <span class="nn">torchvision</span>
|
|
|
|
|
|
<span class="k">class</span> <span class="nc">ConvBlock</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
|
|
<span class="sd">"""Basic convolutional block.</span>
|
|
<span class="sd"> </span>
|
|
<span class="sd"> convolution + batch normalization + relu.</span>
|
|
|
|
<span class="sd"> Args:</span>
|
|
<span class="sd"> in_c (int): number of input channels.</span>
|
|
<span class="sd"> out_c (int): number of output channels.</span>
|
|
<span class="sd"> k (int or tuple): kernel size.</span>
|
|
<span class="sd"> s (int or tuple): stride.</span>
|
|
<span class="sd"> p (int or tuple): padding.</span>
|
|
<span class="sd"> """</span>
|
|
<span class="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_c</span><span class="p">,</span> <span class="n">out_c</span><span class="p">,</span> <span class="n">k</span><span class="p">,</span> <span class="n">s</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">p</span><span class="o">=</span><span class="mi">0</span><span class="p">):</span>
|
|
<span class="nb">super</span><span class="p">(</span><span class="n">ConvBlock</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">conv</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">in_c</span><span class="p">,</span> <span class="n">out_c</span><span class="p">,</span> <span class="n">k</span><span class="p">,</span> <span class="n">stride</span><span class="o">=</span><span class="n">s</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="n">p</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">bn</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="n">out_c</span><span class="p">)</span>
|
|
|
|
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
|
<span class="k">return</span> <span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">bn</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">conv</span><span class="p">(</span><span class="n">x</span><span class="p">)))</span>
|
|
|
|
|
|
<span class="k">class</span> <span class="nc">InceptionA</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
|
|
|
|
<span class="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">,</span> <span class="n">out_channels</span><span class="p">):</span>
|
|
<span class="nb">super</span><span class="p">(</span><span class="n">InceptionA</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
|
<span class="n">mid_channels</span> <span class="o">=</span> <span class="n">out_channels</span> <span class="o">//</span> <span class="mi">4</span>
|
|
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">stream1</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">mid_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">p</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span>
|
|
<span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">stream2</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">mid_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">p</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span>
|
|
<span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">stream3</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">mid_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">p</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span>
|
|
<span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">stream4</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">nn</span><span class="o">.</span><span class="n">AvgPool2d</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span> <span class="n">stride</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
|
|
<span class="p">)</span>
|
|
|
|
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
|
<span class="n">s1</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stream1</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">s2</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stream2</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">s3</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stream3</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">s4</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stream4</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">y</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">([</span><span class="n">s1</span><span class="p">,</span> <span class="n">s2</span><span class="p">,</span> <span class="n">s3</span><span class="p">,</span> <span class="n">s4</span><span class="p">],</span> <span class="n">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">y</span>
|
|
|
|
|
|
<span class="k">class</span> <span class="nc">InceptionB</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
|
|
|
|
<span class="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">,</span> <span class="n">out_channels</span><span class="p">):</span>
|
|
<span class="nb">super</span><span class="p">(</span><span class="n">InceptionB</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
|
<span class="n">mid_channels</span> <span class="o">=</span> <span class="n">out_channels</span> <span class="o">//</span> <span class="mi">4</span>
|
|
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">stream1</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">mid_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">s</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span> <span class="n">p</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span>
|
|
<span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">stream2</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">mid_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">p</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">mid_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">s</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span> <span class="n">p</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span>
|
|
<span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">stream3</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">nn</span><span class="o">.</span><span class="n">MaxPool2d</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span> <span class="n">stride</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span>
|
|
<span class="n">ConvBlock</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">mid_channels</span><span class="o">*</span><span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
|
|
<span class="p">)</span>
|
|
|
|
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
|
<span class="n">s1</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stream1</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">s2</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stream2</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">s3</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stream3</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">y</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">([</span><span class="n">s1</span><span class="p">,</span> <span class="n">s2</span><span class="p">,</span> <span class="n">s3</span><span class="p">],</span> <span class="n">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">y</span>
|
|
|
|
|
|
<span class="k">class</span> <span class="nc">SpatialAttn</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
|
|
<span class="sd">"""Spatial Attention (Sec. 3.1.I.1)"""</span>
|
|
|
|
<span class="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
|
|
<span class="nb">super</span><span class="p">(</span><span class="n">SpatialAttn</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">conv1</span> <span class="o">=</span> <span class="n">ConvBlock</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">s</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span> <span class="n">p</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">conv2</span> <span class="o">=</span> <span class="n">ConvBlock</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
|
|
|
|
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
|
<span class="c1"># global cross-channel averaging</span>
|
|
<span class="n">x</span> <span class="o">=</span> <span class="n">x</span><span class="o">.</span><span class="n">mean</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">keepdim</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
|
|
<span class="c1"># 3-by-3 conv</span>
|
|
<span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv1</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="c1"># bilinear resizing</span>
|
|
<span class="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">upsample</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">2</span><span class="p">)</span><span class="o">*</span><span class="mi">2</span><span class="p">,</span> <span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">3</span><span class="p">)</span><span class="o">*</span><span class="mi">2</span><span class="p">),</span> <span class="n">mode</span><span class="o">=</span><span class="s1">'bilinear'</span><span class="p">,</span> <span class="n">align_corners</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
|
|
<span class="c1"># scaling conv</span>
|
|
<span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv2</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">x</span>
|
|
|
|
|
|
<span class="k">class</span> <span class="nc">ChannelAttn</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
|
|
<span class="sd">"""Channel Attention (Sec. 3.1.I.2)"""</span>
|
|
|
|
<span class="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">,</span> <span class="n">reduction_rate</span><span class="o">=</span><span class="mi">16</span><span class="p">):</span>
|
|
<span class="nb">super</span><span class="p">(</span><span class="n">ChannelAttn</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
|
<span class="k">assert</span> <span class="n">in_channels</span><span class="o">%</span><span class="n">reduction_rate</span> <span class="o">==</span> <span class="mi">0</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">conv1</span> <span class="o">=</span> <span class="n">ConvBlock</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">in_channels</span> <span class="o">//</span> <span class="n">reduction_rate</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">conv2</span> <span class="o">=</span> <span class="n">ConvBlock</span><span class="p">(</span><span class="n">in_channels</span> <span class="o">//</span> <span class="n">reduction_rate</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
|
|
|
|
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
|
<span class="c1"># squeeze operation (global average pooling)</span>
|
|
<span class="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">avg_pool2d</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">()[</span><span class="mi">2</span><span class="p">:])</span>
|
|
<span class="c1"># excitation operation (2 conv layers)</span>
|
|
<span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv1</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv2</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">x</span>
|
|
|
|
|
|
<span class="k">class</span> <span class="nc">SoftAttn</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
|
|
<span class="sd">"""Soft Attention (Sec. 3.1.I)</span>
|
|
<span class="sd"> </span>
|
|
<span class="sd"> Aim: Spatial Attention + Channel Attention</span>
|
|
<span class="sd"> </span>
|
|
<span class="sd"> Output: attention maps with shape identical to input.</span>
|
|
<span class="sd"> """</span>
|
|
|
|
<span class="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">):</span>
|
|
<span class="nb">super</span><span class="p">(</span><span class="n">SoftAttn</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">spatial_attn</span> <span class="o">=</span> <span class="n">SpatialAttn</span><span class="p">()</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">channel_attn</span> <span class="o">=</span> <span class="n">ChannelAttn</span><span class="p">(</span><span class="n">in_channels</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">conv</span> <span class="o">=</span> <span class="n">ConvBlock</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
|
|
|
|
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
|
<span class="n">y_spatial</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">spatial_attn</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">y_channel</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">channel_attn</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">y</span> <span class="o">=</span> <span class="n">y_spatial</span> <span class="o">*</span> <span class="n">y_channel</span>
|
|
<span class="n">y</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">sigmoid</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">conv</span><span class="p">(</span><span class="n">y</span><span class="p">))</span>
|
|
<span class="k">return</span> <span class="n">y</span>
|
|
|
|
|
|
<span class="k">class</span> <span class="nc">HardAttn</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
|
|
<span class="sd">"""Hard Attention (Sec. 3.1.II)"""</span>
|
|
|
|
<span class="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">):</span>
|
|
<span class="nb">super</span><span class="p">(</span><span class="n">HardAttn</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="mi">4</span><span class="o">*</span><span class="mi">2</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">init_params</span><span class="p">()</span>
|
|
|
|
<span class="k">def</span> <span class="nf">init_params</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">zero_</span><span class="p">()</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="o">.</span><span class="n">bias</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">copy_</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([</span><span class="mi">0</span><span class="p">,</span> <span class="o">-</span><span class="mf">0.75</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="o">-</span><span class="mf">0.25</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mf">0.25</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mf">0.75</span><span class="p">],</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">))</span>
|
|
|
|
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
|
<span class="c1"># squeeze operation (global average pooling)</span>
|
|
<span class="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">avg_pool2d</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">()[</span><span class="mi">2</span><span class="p">:])</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">),</span> <span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">1</span><span class="p">))</span>
|
|
<span class="c1"># predict transformation parameters</span>
|
|
<span class="n">theta</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tanh</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="p">(</span><span class="n">x</span><span class="p">))</span>
|
|
<span class="n">theta</span> <span class="o">=</span> <span class="n">theta</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">2</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">theta</span>
|
|
|
|
|
|
<span class="k">class</span> <span class="nc">HarmAttn</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
|
|
<span class="sd">"""Harmonious Attention (Sec. 3.1)"""</span>
|
|
|
|
<span class="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">):</span>
|
|
<span class="nb">super</span><span class="p">(</span><span class="n">HarmAttn</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">soft_attn</span> <span class="o">=</span> <span class="n">SoftAttn</span><span class="p">(</span><span class="n">in_channels</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">hard_attn</span> <span class="o">=</span> <span class="n">HardAttn</span><span class="p">(</span><span class="n">in_channels</span><span class="p">)</span>
|
|
|
|
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
|
<span class="n">y_soft_attn</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">soft_attn</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">theta</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">hard_attn</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">y_soft_attn</span><span class="p">,</span> <span class="n">theta</span>
|
|
|
|
|
|
<div class="viewcode-block" id="HACNN"><a class="viewcode-back" href="../../../pkg/models.html#torchreid.models.hacnn.HACNN">[docs]</a><span class="k">class</span> <span class="nc">HACNN</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
|
|
<span class="sd">"""Harmonious Attention Convolutional Neural Network.</span>
|
|
|
|
<span class="sd"> Reference:</span>
|
|
<span class="sd"> Li et al. Harmonious Attention Network for Person Re-identification. CVPR 2018.</span>
|
|
|
|
<span class="sd"> Public keys:</span>
|
|
<span class="sd"> - ``hacnn``: HACNN.</span>
|
|
<span class="sd"> """</span>
|
|
|
|
<span class="c1"># Args:</span>
|
|
<span class="c1"># num_classes (int): number of classes to predict</span>
|
|
<span class="c1"># nchannels (list): number of channels AFTER concatenation</span>
|
|
<span class="c1"># feat_dim (int): feature dimension for a single stream</span>
|
|
<span class="c1"># learn_region (bool): whether to learn region features (i.e. local branch)</span>
|
|
|
|
<span class="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">num_classes</span><span class="p">,</span> <span class="n">loss</span><span class="o">=</span><span class="s1">'softmax'</span><span class="p">,</span> <span class="n">nchannels</span><span class="o">=</span><span class="p">[</span><span class="mi">128</span><span class="p">,</span> <span class="mi">256</span><span class="p">,</span> <span class="mi">384</span><span class="p">],</span> <span class="n">feat_dim</span><span class="o">=</span><span class="mi">512</span><span class="p">,</span> <span class="n">learn_region</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">use_gpu</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
|
|
<span class="nb">super</span><span class="p">(</span><span class="n">HACNN</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">loss</span> <span class="o">=</span> <span class="n">loss</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">learn_region</span> <span class="o">=</span> <span class="n">learn_region</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">use_gpu</span> <span class="o">=</span> <span class="n">use_gpu</span>
|
|
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">conv</span> <span class="o">=</span> <span class="n">ConvBlock</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">s</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span> <span class="n">p</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
|
|
|
|
<span class="c1"># Construct Inception + HarmAttn blocks</span>
|
|
<span class="c1"># ============== Block 1 ==============</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">inception1</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">InceptionA</span><span class="p">(</span><span class="mi">32</span><span class="p">,</span> <span class="n">nchannels</span><span class="p">[</span><span class="mi">0</span><span class="p">]),</span>
|
|
<span class="n">InceptionB</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">nchannels</span><span class="p">[</span><span class="mi">0</span><span class="p">]),</span>
|
|
<span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">ha1</span> <span class="o">=</span> <span class="n">HarmAttn</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span>
|
|
|
|
<span class="c1"># ============== Block 2 ==============</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">inception2</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">InceptionA</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">nchannels</span><span class="p">[</span><span class="mi">1</span><span class="p">]),</span>
|
|
<span class="n">InceptionB</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span> <span class="n">nchannels</span><span class="p">[</span><span class="mi">1</span><span class="p">]),</span>
|
|
<span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">ha2</span> <span class="o">=</span> <span class="n">HarmAttn</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">1</span><span class="p">])</span>
|
|
|
|
<span class="c1"># ============== Block 3 ==============</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">inception3</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">InceptionA</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span> <span class="n">nchannels</span><span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
|
|
<span class="n">InceptionB</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">2</span><span class="p">],</span> <span class="n">nchannels</span><span class="p">[</span><span class="mi">2</span><span class="p">]),</span>
|
|
<span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">ha3</span> <span class="o">=</span> <span class="n">HarmAttn</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">2</span><span class="p">])</span>
|
|
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc_global</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">2</span><span class="p">],</span> <span class="n">feat_dim</span><span class="p">),</span>
|
|
<span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm1d</span><span class="p">(</span><span class="n">feat_dim</span><span class="p">),</span>
|
|
<span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(),</span>
|
|
<span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">classifier_global</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">feat_dim</span><span class="p">,</span> <span class="n">num_classes</span><span class="p">)</span>
|
|
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">learn_region</span><span class="p">:</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">init_scale_factors</span><span class="p">()</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">local_conv1</span> <span class="o">=</span> <span class="n">InceptionB</span><span class="p">(</span><span class="mi">32</span><span class="p">,</span> <span class="n">nchannels</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">local_conv2</span> <span class="o">=</span> <span class="n">InceptionB</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">nchannels</span><span class="p">[</span><span class="mi">1</span><span class="p">])</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">local_conv3</span> <span class="o">=</span> <span class="n">InceptionB</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span> <span class="n">nchannels</span><span class="p">[</span><span class="mi">2</span><span class="p">])</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">fc_local</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
|
<span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">nchannels</span><span class="p">[</span><span class="mi">2</span><span class="p">]</span><span class="o">*</span><span class="mi">4</span><span class="p">,</span> <span class="n">feat_dim</span><span class="p">),</span>
|
|
<span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm1d</span><span class="p">(</span><span class="n">feat_dim</span><span class="p">),</span>
|
|
<span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(),</span>
|
|
<span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">classifier_local</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">feat_dim</span><span class="p">,</span> <span class="n">num_classes</span><span class="p">)</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">feat_dim</span> <span class="o">=</span> <span class="n">feat_dim</span> <span class="o">*</span> <span class="mi">2</span>
|
|
<span class="k">else</span><span class="p">:</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">feat_dim</span> <span class="o">=</span> <span class="n">feat_dim</span>
|
|
|
|
<span class="k">def</span> <span class="nf">init_scale_factors</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
|
|
<span class="c1"># initialize scale factors (s_w, s_h) for four regions</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">scale_factors</span> <span class="o">=</span> <span class="p">[]</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">scale_factors</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mf">0.25</span><span class="p">]],</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">))</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">scale_factors</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mf">0.25</span><span class="p">]],</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">))</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">scale_factors</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mf">0.25</span><span class="p">]],</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">))</span>
|
|
<span class="bp">self</span><span class="o">.</span><span class="n">scale_factors</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mf">0.25</span><span class="p">]],</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">))</span>
|
|
|
|
<span class="k">def</span> <span class="nf">stn</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">,</span> <span class="n">theta</span><span class="p">):</span>
|
|
<span class="sd">"""Performs spatial transform</span>
|
|
<span class="sd"> </span>
|
|
<span class="sd"> x: (batch, channel, height, width)</span>
|
|
<span class="sd"> theta: (batch, 2, 3)</span>
|
|
<span class="sd"> """</span>
|
|
<span class="n">grid</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">affine_grid</span><span class="p">(</span><span class="n">theta</span><span class="p">,</span> <span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">())</span>
|
|
<span class="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">grid_sample</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">grid</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">x</span>
|
|
|
|
<span class="k">def</span> <span class="nf">transform_theta</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">theta_i</span><span class="p">,</span> <span class="n">region_idx</span><span class="p">):</span>
|
|
<span class="sd">"""Transforms theta to include (s_w, s_h), resulting in (batch, 2, 3)"""</span>
|
|
<span class="n">scale_factors</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">scale_factors</span><span class="p">[</span><span class="n">region_idx</span><span class="p">]</span>
|
|
<span class="n">theta</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="n">theta_i</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">),</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
|
|
<span class="n">theta</span><span class="p">[:,:,:</span><span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="n">scale_factors</span>
|
|
<span class="n">theta</span><span class="p">[:,:,</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span> <span class="o">=</span> <span class="n">theta_i</span>
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">use_gpu</span><span class="p">:</span> <span class="n">theta</span> <span class="o">=</span> <span class="n">theta</span><span class="o">.</span><span class="n">cuda</span><span class="p">()</span>
|
|
<span class="k">return</span> <span class="n">theta</span>
|
|
|
|
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
|
<span class="k">assert</span> <span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">2</span><span class="p">)</span> <span class="o">==</span> <span class="mi">160</span> <span class="ow">and</span> <span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">3</span><span class="p">)</span> <span class="o">==</span> <span class="mi">64</span><span class="p">,</span> \
|
|
<span class="s1">'Input size does not match, expected (160, 64) but got (</span><span class="si">{}</span><span class="s1">, </span><span class="si">{}</span><span class="s1">)'</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">2</span><span class="p">),</span> <span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">3</span><span class="p">))</span>
|
|
<span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
|
|
<span class="c1"># ============== Block 1 ==============</span>
|
|
<span class="c1"># global branch</span>
|
|
<span class="n">x1</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">inception1</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
|
<span class="n">x1_attn</span><span class="p">,</span> <span class="n">x1_theta</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">ha1</span><span class="p">(</span><span class="n">x1</span><span class="p">)</span>
|
|
<span class="n">x1_out</span> <span class="o">=</span> <span class="n">x1</span> <span class="o">*</span> <span class="n">x1_attn</span>
|
|
<span class="c1"># local branch</span>
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">learn_region</span><span class="p">:</span>
|
|
<span class="n">x1_local_list</span> <span class="o">=</span> <span class="p">[]</span>
|
|
<span class="k">for</span> <span class="n">region_idx</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="mi">4</span><span class="p">):</span>
|
|
<span class="n">x1_theta_i</span> <span class="o">=</span> <span class="n">x1_theta</span><span class="p">[:,</span><span class="n">region_idx</span><span class="p">,:]</span>
|
|
<span class="n">x1_theta_i</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">transform_theta</span><span class="p">(</span><span class="n">x1_theta_i</span><span class="p">,</span> <span class="n">region_idx</span><span class="p">)</span>
|
|
<span class="n">x1_trans_i</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stn</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">x1_theta_i</span><span class="p">)</span>
|
|
<span class="n">x1_trans_i</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">upsample</span><span class="p">(</span><span class="n">x1_trans_i</span><span class="p">,</span> <span class="p">(</span><span class="mi">24</span><span class="p">,</span> <span class="mi">28</span><span class="p">),</span> <span class="n">mode</span><span class="o">=</span><span class="s1">'bilinear'</span><span class="p">,</span> <span class="n">align_corners</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
|
|
<span class="n">x1_local_i</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">local_conv1</span><span class="p">(</span><span class="n">x1_trans_i</span><span class="p">)</span>
|
|
<span class="n">x1_local_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">x1_local_i</span><span class="p">)</span>
|
|
|
|
<span class="c1"># ============== Block 2 ==============</span>
|
|
<span class="c1"># Block 2</span>
|
|
<span class="c1"># global branch</span>
|
|
<span class="n">x2</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">inception2</span><span class="p">(</span><span class="n">x1_out</span><span class="p">)</span>
|
|
<span class="n">x2_attn</span><span class="p">,</span> <span class="n">x2_theta</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">ha2</span><span class="p">(</span><span class="n">x2</span><span class="p">)</span>
|
|
<span class="n">x2_out</span> <span class="o">=</span> <span class="n">x2</span> <span class="o">*</span> <span class="n">x2_attn</span>
|
|
<span class="c1"># local branch</span>
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">learn_region</span><span class="p">:</span>
|
|
<span class="n">x2_local_list</span> <span class="o">=</span> <span class="p">[]</span>
|
|
<span class="k">for</span> <span class="n">region_idx</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="mi">4</span><span class="p">):</span>
|
|
<span class="n">x2_theta_i</span> <span class="o">=</span> <span class="n">x2_theta</span><span class="p">[:,</span><span class="n">region_idx</span><span class="p">,:]</span>
|
|
<span class="n">x2_theta_i</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">transform_theta</span><span class="p">(</span><span class="n">x2_theta_i</span><span class="p">,</span> <span class="n">region_idx</span><span class="p">)</span>
|
|
<span class="n">x2_trans_i</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stn</span><span class="p">(</span><span class="n">x1_out</span><span class="p">,</span> <span class="n">x2_theta_i</span><span class="p">)</span>
|
|
<span class="n">x2_trans_i</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">upsample</span><span class="p">(</span><span class="n">x2_trans_i</span><span class="p">,</span> <span class="p">(</span><span class="mi">12</span><span class="p">,</span> <span class="mi">14</span><span class="p">),</span> <span class="n">mode</span><span class="o">=</span><span class="s1">'bilinear'</span><span class="p">,</span> <span class="n">align_corners</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
|
|
<span class="n">x2_local_i</span> <span class="o">=</span> <span class="n">x2_trans_i</span> <span class="o">+</span> <span class="n">x1_local_list</span><span class="p">[</span><span class="n">region_idx</span><span class="p">]</span>
|
|
<span class="n">x2_local_i</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">local_conv2</span><span class="p">(</span><span class="n">x2_local_i</span><span class="p">)</span>
|
|
<span class="n">x2_local_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">x2_local_i</span><span class="p">)</span>
|
|
|
|
<span class="c1"># ============== Block 3 ==============</span>
|
|
<span class="c1"># Block 3</span>
|
|
<span class="c1"># global branch</span>
|
|
<span class="n">x3</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">inception3</span><span class="p">(</span><span class="n">x2_out</span><span class="p">)</span>
|
|
<span class="n">x3_attn</span><span class="p">,</span> <span class="n">x3_theta</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">ha3</span><span class="p">(</span><span class="n">x3</span><span class="p">)</span>
|
|
<span class="n">x3_out</span> <span class="o">=</span> <span class="n">x3</span> <span class="o">*</span> <span class="n">x3_attn</span>
|
|
<span class="c1"># local branch</span>
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">learn_region</span><span class="p">:</span>
|
|
<span class="n">x3_local_list</span> <span class="o">=</span> <span class="p">[]</span>
|
|
<span class="k">for</span> <span class="n">region_idx</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="mi">4</span><span class="p">):</span>
|
|
<span class="n">x3_theta_i</span> <span class="o">=</span> <span class="n">x3_theta</span><span class="p">[:,</span><span class="n">region_idx</span><span class="p">,:]</span>
|
|
<span class="n">x3_theta_i</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">transform_theta</span><span class="p">(</span><span class="n">x3_theta_i</span><span class="p">,</span> <span class="n">region_idx</span><span class="p">)</span>
|
|
<span class="n">x3_trans_i</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stn</span><span class="p">(</span><span class="n">x2_out</span><span class="p">,</span> <span class="n">x3_theta_i</span><span class="p">)</span>
|
|
<span class="n">x3_trans_i</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">upsample</span><span class="p">(</span><span class="n">x3_trans_i</span><span class="p">,</span> <span class="p">(</span><span class="mi">6</span><span class="p">,</span> <span class="mi">7</span><span class="p">),</span> <span class="n">mode</span><span class="o">=</span><span class="s1">'bilinear'</span><span class="p">,</span> <span class="n">align_corners</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
|
|
<span class="n">x3_local_i</span> <span class="o">=</span> <span class="n">x3_trans_i</span> <span class="o">+</span> <span class="n">x2_local_list</span><span class="p">[</span><span class="n">region_idx</span><span class="p">]</span>
|
|
<span class="n">x3_local_i</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">local_conv3</span><span class="p">(</span><span class="n">x3_local_i</span><span class="p">)</span>
|
|
<span class="n">x3_local_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">x3_local_i</span><span class="p">)</span>
|
|
|
|
<span class="c1"># ============== Feature generation ==============</span>
|
|
<span class="c1"># global branch</span>
|
|
<span class="n">x_global</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">avg_pool2d</span><span class="p">(</span><span class="n">x3_out</span><span class="p">,</span> <span class="n">x3_out</span><span class="o">.</span><span class="n">size</span><span class="p">()[</span><span class="mi">2</span><span class="p">:])</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">x3_out</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">),</span> <span class="n">x3_out</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">1</span><span class="p">))</span>
|
|
<span class="n">x_global</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc_global</span><span class="p">(</span><span class="n">x_global</span><span class="p">)</span>
|
|
<span class="c1"># local branch</span>
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">learn_region</span><span class="p">:</span>
|
|
<span class="n">x_local_list</span> <span class="o">=</span> <span class="p">[]</span>
|
|
<span class="k">for</span> <span class="n">region_idx</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="mi">4</span><span class="p">):</span>
|
|
<span class="n">x_local_i</span> <span class="o">=</span> <span class="n">x3_local_list</span><span class="p">[</span><span class="n">region_idx</span><span class="p">]</span>
|
|
<span class="n">x_local_i</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">avg_pool2d</span><span class="p">(</span><span class="n">x_local_i</span><span class="p">,</span> <span class="n">x_local_i</span><span class="o">.</span><span class="n">size</span><span class="p">()[</span><span class="mi">2</span><span class="p">:])</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">x_local_i</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">),</span> <span class="o">-</span><span class="mi">1</span><span class="p">)</span>
|
|
<span class="n">x_local_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">x_local_i</span><span class="p">)</span>
|
|
<span class="n">x_local</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">(</span><span class="n">x_local_list</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
|
|
<span class="n">x_local</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc_local</span><span class="p">(</span><span class="n">x_local</span><span class="p">)</span>
|
|
|
|
<span class="k">if</span> <span class="ow">not</span> <span class="bp">self</span><span class="o">.</span><span class="n">training</span><span class="p">:</span>
|
|
<span class="c1"># l2 normalization before concatenation</span>
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">learn_region</span><span class="p">:</span>
|
|
<span class="n">x_global</span> <span class="o">=</span> <span class="n">x_global</span> <span class="o">/</span> <span class="n">x_global</span><span class="o">.</span><span class="n">norm</span><span class="p">(</span><span class="n">p</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span> <span class="n">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">keepdim</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
|
|
<span class="n">x_local</span> <span class="o">=</span> <span class="n">x_local</span> <span class="o">/</span> <span class="n">x_local</span><span class="o">.</span><span class="n">norm</span><span class="p">(</span><span class="n">p</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span> <span class="n">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">keepdim</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
|
|
<span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">([</span><span class="n">x_global</span><span class="p">,</span> <span class="n">x_local</span><span class="p">],</span> <span class="mi">1</span><span class="p">)</span>
|
|
<span class="k">else</span><span class="p">:</span>
|
|
<span class="k">return</span> <span class="n">x_global</span>
|
|
|
|
<span class="n">prelogits_global</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">classifier_global</span><span class="p">(</span><span class="n">x_global</span><span class="p">)</span>
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">learn_region</span><span class="p">:</span>
|
|
<span class="n">prelogits_local</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">classifier_local</span><span class="p">(</span><span class="n">x_local</span><span class="p">)</span>
|
|
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">loss</span> <span class="o">==</span> <span class="s1">'softmax'</span><span class="p">:</span>
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">learn_region</span><span class="p">:</span>
|
|
<span class="k">return</span> <span class="p">(</span><span class="n">prelogits_global</span><span class="p">,</span> <span class="n">prelogits_local</span><span class="p">)</span>
|
|
<span class="k">else</span><span class="p">:</span>
|
|
<span class="k">return</span> <span class="n">prelogits_global</span>
|
|
|
|
<span class="k">elif</span> <span class="bp">self</span><span class="o">.</span><span class="n">loss</span> <span class="o">==</span> <span class="s1">'triplet'</span><span class="p">:</span>
|
|
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">learn_region</span><span class="p">:</span>
|
|
<span class="k">return</span> <span class="p">(</span><span class="n">prelogits_global</span><span class="p">,</span> <span class="n">prelogits_local</span><span class="p">),</span> <span class="p">(</span><span class="n">x_global</span><span class="p">,</span> <span class="n">x_local</span><span class="p">)</span>
|
|
<span class="k">else</span><span class="p">:</span>
|
|
<span class="k">return</span> <span class="n">prelogits_global</span><span class="p">,</span> <span class="n">x_global</span>
|
|
|
|
<span class="k">else</span><span class="p">:</span>
|
|
<span class="k">raise</span> <span class="ne">KeyError</span><span class="p">(</span><span class="s2">"Unsupported loss: </span><span class="si">{}</span><span class="s2">"</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">loss</span><span class="p">))</span></div>
|
|
</pre></div>
|
|
|
|
</div>
|
|
|
|
</div>
|
|
<footer>
|
|
|
|
|
|
<hr/>
|
|
|
|
<div role="contentinfo">
|
|
<p>
|
|
© Copyright 2019, Kaiyang Zhou
|
|
|
|
</p>
|
|
</div>
|
|
Built with <a href="http://sphinx-doc.org/">Sphinx</a> using a <a href="https://github.com/rtfd/sphinx_rtd_theme">theme</a> provided by <a href="https://readthedocs.org">Read the Docs</a>.
|
|
|
|
</footer>
|
|
|
|
</div>
|
|
</div>
|
|
|
|
</section>
|
|
|
|
</div>
|
|
|
|
|
|
|
|
<script type="text/javascript">
|
|
jQuery(function () {
|
|
SphinxRtdTheme.Navigation.enable(true);
|
|
});
|
|
</script>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
</body>
|
|
</html> |