---

title: Additional Information


keywords: fastai
sidebar: home_sidebar

summary: "This Notebook contains information on use of _deepflash2_."
description: "This Notebook contains information on use of _deepflash2_."
nb_path: "nbs/add_information.ipynb"
---
<!--

#################################################
### THIS FILE WAS AUTOGENERATED! DO NOT EDIT! ###
#################################################
# file to edit: nbs/add_information.ipynb
# command to build the docs after a change: nbdev_build_docs

-->

<div class="container" id="notebook-container">
        
    {% raw %}
    
<div class="cell border-box-sizing code_cell rendered">

</div>
    {% endraw %}

<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<h2 id="Naming">Naming<a class="anchor-link" href="#Naming"> </a></h2>
</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<ul>
<li>Images must have unique name or ID<ul>
<li>_0001.tif --&gt; name/ID: 0001; img_5.png --&gt; name/ID: img<em>5, ...</em> </li>
</ul>
</li>
<li>Corresponding masks must start with name or ID + a mask suffix<ul>
<li>_0001 -&gt; 0001_mask.png (mask_suffix = "<em>mask.png")</em></li>
<li>_0001 -&gt; 0001.png (mask<em>suffix = ".png")</em></li>
</ul>
</li>
</ul>

</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<h2 id="Learning-Rate-Finder">Learning-Rate-Finder<a class="anchor-link" href="#Learning-Rate-Finder"> </a></h2><ul>
<li><strong>Learning rate</strong>: controls how quickly or slowly a neural network model learns. <br>
We need to set the <strong>maximum learning rate</strong> <code>max_lr</code> when traing with the <strong><a href="https://fastai1.fast.ai/callbacks.one_cycle.html#The-1cycle-policy">One-Cycle-Policy</a></strong>. Using the <strong>Learning Rate Finder</strong> below, a good value for <code>max_lr</code> is somthing in the range of:<ul>
<li>one tenth of the minimum before the divergence (<em>Minimum/10</em>)</li>
<li>when the slope is the steepest (<em>steepest point</em>) </li>
<li>In our experiments, we found that a <strong>maximum learning rate of 5e-4</strong> (e.g., 0.0005) yielded the best results across experiments.</li>
</ul>
</li>
</ul>

</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<h2 id="Train-validation-split">Train-validation-split<a class="anchor-link" href="#Train-validation-split"> </a></h2><ul>
<li><code>n_models = 1</code> leads to a train-validation-split of 0.75/0.25</li>
<li><code>n_models &gt; 1</code> (model ensembles) to _<a href="https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.KFold.html">k-fold cross validation</a>_</li>
</ul>

</div>
</div>
</div>
</div>