How to create a transfer learning model
<p>In <strong>convolutional neural networks,</strong> to reach the best accuracy and generalization ability, we need to test many experiments, adding layers or hidden units and so on. Sometimes it will take a lot of time and it can be just a random guess.</p>
<p>But one thing is for sure, <strong>neural network</strong> is not a brand-new thing. Surely someone must have done the work and built a workable neural network model for specific problems. And there it is!</p>
<p>So what if we use the model trained by other professionals and apply it on our own problem? That is what <strong>Transfer Learning</strong> does.</p>
<p> </p>
<p><em>Illustration from (zero to mastery course)</em></p>
<p>Transfer learning can help us:</p>
<ul>
<li>Leverage an existing trained model that has been proven to work — save time</li>
<li>Leverage a model that has already learned patterns on problems that we’re working on.</li>
</ul>
<p>How to find available model structures and how to use them? Here I introduce two methods. In this blog, I take <strong>image domain</strong> and <strong>EfficientNetB0</strong> as an example of architecture.</p>
<p>Before we start, let’s assume that <strong>we are solving a computer vision problem</strong>.</p>
<h1>Download the model from TensorFlow Hub</h1>
<p>First, let’s go to the [<a href="https://tfhub.dev/" rel="noopener ugc nofollow" target="_blank">TensorFlow Hub</a>]. On the homepage, we can see different types of problem domains.</p>
<p><a href="https://medium.com/@719458145/how-to-create-a-transfer-learning-model-two-methods-1b9441a385e7">Website</a></p>