-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathinfo.html
More file actions
114 lines (90 loc) · 6.55 KB
/
Copy pathinfo.html
File metadata and controls
114 lines (90 loc) · 6.55 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" /><meta name="generator" content="Docutils 0.19: https://docutils.sourceforge.io/" />
<title>DRAGen — RST: Decision Region Analysis documentation</title>
<link rel="stylesheet" type="text/css" href="_static/pygments.css" />
<link rel="stylesheet" type="text/css" href="_static/alabaster.css" />
<script data-url_root="./" id="documentation_options" src="_static/documentation_options.js"></script>
<script src="_static/doctools.js"></script>
<script src="_static/sphinx_highlight.js"></script>
<link rel="index" title="Index" href="genindex.html" />
<link rel="search" title="Search" href="search.html" />
<link rel="stylesheet" href="_static/custom.css" type="text/css" />
<meta name="viewport" content="width=device-width, initial-scale=0.9, maximum-scale=0.9" />
</head><body>
<div class="document">
<div class="documentwrapper">
<div class="bodywrapper">
<div class="body" role="main">
<section id="dragen">
<h1>DRAGen<a class="headerlink" href="#dragen" title="Permalink to this heading">¶</a></h1>
<p>Decsion Region Analysis for Generalizability (DRAGen) is a tool to analyze the decision space of an image classification model to increase understanding of the model’s generalizability. A model’s decision space maps a change in the input to a change in the model’s output. DRAGen utilizes triplets of image samples to generate vicinal distributions of virtual images, created by linearly interpolating between the triplet images. These virtual images increase the density of available samples in the decision space, which allows for characterization of the decision space beyond the original finite data set. This insight into the decision space composition indicates how the model is likely to behave on data distributions upon which the model cannot generalize well.</p>
</section>
<section id="getting-started">
<h1>Getting Started<a class="headerlink" href="#getting-started" title="Permalink to this heading">¶</a></h1>
<p>Three inputs are required for this RST:</p>
<ol class="arabic simple">
<li><p><strong>Trained Model:</strong> the model must be saved in onnx format. Information on how to convert your model to onnx format can be found at the <a class="reference external" href="https://github.com/onnx/tutorials#converting-to-onnx-format">onnx GitHub page</a>.</p></li>
<li><p><strong>Images:</strong> Images can be saved in any format supported by PIL.</p></li>
<li><p><strong>Input csv:</strong> A csv file which can be used to map the image paths to subgroup attributes.</p></li>
</ol>
<p>Examples implementation can be found in the <a class="reference external" href="https://github.com/DIDSR/RST_Decision_Region_Analysis/tree/main/examples">examples folder</a>.</p>
<p>Scripts to generate and analyze decision regions can be found in the <code class="docutils literal notranslate"><span class="pre">test</span></code> folder,
all arguments used to run these scripts are located in <code class="docutils literal notranslate"><span class="pre">src/args.py</span></code>.
Example inputs are included in the <code class="docutils literal notranslate"><span class="pre">examples</span></code> folder.</p>
</section>
<section id="terminology">
<h1>Terminology<a class="headerlink" href="#terminology" title="Permalink to this heading">¶</a></h1>
<ul class="simple">
<li><p><code class="docutils literal notranslate"><span class="pre">class</span></code>: An attribute by which the model classifies images. Only binary classification models are currently supported.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">subgroup</span> <span class="pre">attribute</span></code>: An attribute by which the model <em>does not</em> classify images, but can be used to group samples into subgroups.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">decision</span> <span class="pre">region</span></code>: A portion of the decision space. The decision regions generated in this RST are the regions of the decision space near to a ‘triplet’ of sample images.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">virtual</span> <span class="pre">image</span></code>: An image that was created by modifying existing image(s), rather than obtained through a typical image acquisition method.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">vicinal</span> <span class="pre">distribution</span></code>: The collection of virtual images created by linearly interpolating between a ‘triplet’ of three images.</p></li>
</ul>
</section>
</div>
</div>
</div>
<div class="sphinxsidebar" role="navigation" aria-label="main navigation">
<div class="sphinxsidebarwrapper">
<div id="searchbox" style="display: none" role="search">
<h3 id="searchlabel">Quick search</h3>
<div class="searchformwrapper">
<form class="search" action="search.html" method="get">
<input type="text" name="q" aria-labelledby="searchlabel" autocomplete="off" autocorrect="off" autocapitalize="off" spellcheck="false"/>
<input type="submit" value="Go" />
</form>
</div>
</div>
<script>document.getElementById('searchbox').style.display = "block"</script>
<h3><a href="index.html">Table of Contents</a></h3>
<ul>
<li class="toctree-l1"><a class="reference internal" href="index.html">DRAGen</a></li>
<li class="toctree-l1"><a class="reference internal" href="src.html">Data Input</a></li>
<li class="toctree-l1"><a class="reference internal" href="src.html#decision-region-generation">Decision Region Generation</a></li>
<li class="toctree-l1"><a class="reference internal" href="src.html#module-src.composition_analysis">Composition Analysis</a></li>
<li class="toctree-l1"><a class="reference internal" href="args.html">Command Line Arguments</a></li>
</ul>
<div role="note" aria-label="source link">
<h3>This Page</h3>
<ul class="this-page-menu">
<li><a href="_sources/info.rst.txt"
rel="nofollow">Show Source</a></li>
</ul>
</div>
</div>
</div>
<div class="clearer"></div>
</div>
<div class="footer">
Powered by <a href="http://sphinx-doc.org/">Sphinx 6.1.3</a>
& <a href="https://github.com/bitprophet/alabaster">Alabaster 0.7.13</a>
|
<a href="_sources/info.rst.txt"
rel="nofollow">Page source</a>
</div>
</body>
</html>