Deployed 59f87ee with MkDocs version: 1.6.1

Choosing-the-Model/index.html

@@ -303,6 +303,38 @@
     </span>
 </a>
 </li>
+<li class="md-nav__item">
+<a class="md-nav__link" href="#model-naming-convention">1.4 
+<span class="md-ellipsis">
+      Model Naming Convention
+    </span>
+</a>
+<nav aria-label="Model Naming Convention" class="md-nav">
+<ul class="md-nav__list">
+<li class="md-nav__item">
+<a class="md-nav__link" href="#version-number">1.4.1 
+<span class="md-ellipsis">
+      Version Number
+    </span>
+</a>
+</li>
+<li class="md-nav__item">
+<a class="md-nav__link" href="#size-indicator">1.4.2 
+<span class="md-ellipsis">
+      Size Indicator
+    </span>
+</a>
+</li>
+<li class="md-nav__item">
+<a class="md-nav__link" href="#quantization-indicator">1.4.3 
+<span class="md-ellipsis">
+      Quantization Indicator
+    </span>
+</a>
+</li>
+</ul>
+</nav>
+</li>
 </ul>
 </nav>
 </li>
@@ -761,6 +793,38 @@
     </span>
 </a>
 </li>
+<li class="md-nav__item">
+<a class="md-nav__link" href="#model-naming-convention">1.4 
+<span class="md-ellipsis">
+      Model Naming Convention
+    </span>
+</a>
+<nav aria-label="Model Naming Convention" class="md-nav">
+<ul class="md-nav__list">
+<li class="md-nav__item">
+<a class="md-nav__link" href="#version-number">1.4.1 
+<span class="md-ellipsis">
+      Version Number
+    </span>
+</a>
+</li>
+<li class="md-nav__item">
+<a class="md-nav__link" href="#size-indicator">1.4.2 
+<span class="md-ellipsis">
+      Size Indicator
+    </span>
+</a>
+</li>
+<li class="md-nav__item">
+<a class="md-nav__link" href="#quantization-indicator">1.4.3 
+<span class="md-ellipsis">
+      Quantization Indicator
+    </span>
+</a>
+</li>
+</ul>
+</nav>
+</li>
 </ul>
 </nav>
 </div>
@@ -782,7 +846,8 @@ For more technical/deeper explanations have a look on <a href="../Neural-Network
 <li><code>dig-class11</code></li>
 <li><code>dig-class100</code> and <code>dig-cont</code></li>
 </ul>
-<p><code>class</code> stands for <strong>classification</strong> and <code>cont</code> stands for <strong>continuous</strong>. The <code>11</code> means that there are 11 states (<code>0..9</code> and <code>N</code>). The <code>100</code> indicates that the model resolves into <code>x.1</code> steps.</p>
+<p><code>class</code> stands for <strong>classification</strong> and <code>cont</code> stands for <strong>continuous</strong>. The <code>11</code> means that there are 11 states (<code>0..9</code> and <code>N</code>). The <code>100</code> indicates that the model resolves into <code>x.1</code> steps by having 100 states (0.0, 0.1, 0.2, ... 9.7, 9.8, 9.9).</p>
+<p><strong>continuous</strong> means, that there is no discrete model, that has discreate states, but there is a different mechanism, that provides a not discrete value in the interval between [0, 1[.</p>
 <h3 id="dig-class11"><span class="enumerate-headings-plugin enumerate-heading-plugin">1.1.1</span> <code>dig-class11</code></h3>
 <p>This model can recognize <strong>full digits</strong>. It was the first model version. All intermediate states shown a <code>N</code> for not-a-number (aka. <code>NaN</code>). But in post process it uses older values to fill up the <code>N</code> values if possible.</p>
 <p><img alt="" src="../img/dig-class11.png" style="width:300px"/></p>
@@ -853,6 +918,37 @@ Usually the models are distrusted therefore in both versions. They can be distin
 </tr>
 </tbody>
 </table>
+<h2 id="model-naming-convention"><span class="enumerate-headings-plugin enumerate-heading-plugin">1.4</span> Model Naming Convention</h2>
+<p>Model filenames follow a specific structure composed of several parts:</p>
+<ol>
+<li><strong>Model Type</strong>: Indicates the kind of model, such as <code>ana/dig</code>, <code>cont</code>, <code>class11</code>, or <code>class100</code>.</li>
+<li><strong>Version Number</strong>: Denotes the version of the model.</li>
+<li><strong>Size Indicator</strong>: Represents the size or complexity of the model.</li>
+<li><strong>Quantization Indicator</strong> (optional): Specifies whether the model was quantized after training.</li>
+<li><strong>File Type</strong>: Always <code>.tflite</code>.</li>
+</ol>
+<h3 id="version-number"><span class="enumerate-headings-plugin enumerate-heading-plugin">1.4.1</span> Version Number</h3>
+<p>The version number consists of four digits:</p>
+<ul>
+<li>The <strong>first two digits</strong> represent the <em>main version</em>.</li>
+<li>The <strong>last two digits</strong> represent the <em>sub-version</em>.</li>
+</ul>
+<p>Example:
+ <code>0610</code> → Main version: <code>06</code>, Sub-version: <code>10</code>
+ In general, higher numbers correspond to newer models.</p>
+<h3 id="size-indicator"><span class="enumerate-headings-plugin enumerate-heading-plugin">1.4.2</span> Size Indicator</h3>
+<ul>
+<li>The size indicator typically starts with <code>s</code> followed by a number, e.g., <code>s3</code>.</li>
+<li>This naming is not strictly standardized yet.</li>
+<li>Training usually begins with a larger model (<code>s0</code>), and successive versions (<code>s1</code>, <code>s2</code>, ...) reduce the number of parameters to create smaller, faster networks.</li>
+<li>Reducing model size can improve efficiency, but may eventually lead to a drop in recognition accuracy.</li>
+<li>In newer model generations, only the best-performing sizes are retained and further trained.</li>
+</ul>
+<h3 id="quantization-indicator"><span class="enumerate-headings-plugin enumerate-heading-plugin">1.4.3</span> Quantization Indicator</h3>
+<ul>
+<li>If the model name ends with <code>-q</code>, it indicates that <strong>quantization</strong> was applied after training.</li>
+<li>Quantization significantly reduces the model size, typically without a noticeable impact on recognition performance.</li>
+</ul>
 </article>
 </div>
 <script>var target=document.getElementById(location.hash.slice(1));target&&target.name&&(target.checked=target.name.startsWith("__tabbed_"))</script>