Gain insights into image data processing and CNNs for image classification using TensorFlow. Delve into CNN architectures and their applications, requiring Python and TensorFlow knowledge.

irml.tar.gz

image-rec

If you want to dive into the technology behind computer vision and self driving cars, this is where to start.

In this course you'll learn how to process data from image files and create convolutional neural networks (CNNs) to classify different types of images. After completing this course, you will have a solid understanding of why CNNs work and which CNN architectures to use for different tasks.

The code for this course is built around the TensorFlow framework, one of the premier frameworks for industry machine learning, and the Python pandas library for data analysis. Knowledge of Python and TensorFlow are prerequisites. 

This course was created by AdaptiLab, a company specializing in evaluating, sourcing, and upskilling enterprise machine learning talent. It is built in collaboration with industry machine learning experts from Google, Microsoft, Amazon, and Apple.

Image Recognition with Machine Learning

<h2 class="section-title">Chapter Goals:</h2>
<ul>
	<li>Understand the purpose of residual learning</li>
	<li>Learn the structure of a ResNet building block</li>
	<li>Create a function for a ResNet building block</li>
</ul>


<h2>A. Learning the identity</h2>

In the previous chapter, we touched upon <i>residual learning</i>, i.e. the process of a block learning the <i>residual function</i>:
$$
\small
\mathcal{F}_B(\mathbf{x}) = \mathcal{H}_B(\mathbf{x}) - \mathbf{x}
$$
where $\small \mathcal{H}_B(\mathbf{x})$ is the optimal mapping function.</p>
<p>You might ask why we use residual learning, since the residual function is pretty similar to the optimal function. The reason is because it is much easier for a block to learn a <i>zero mapping</i> than an identity mapping.</p>

A zero mapping simply returns an output of all 0's. This is very easy to learn; the block can just be trained so that all its weights are 0. However, if the block learns a zero mapping, i.e. $\small \mathcal{F}_B(\mathbf{x}) = \mathbf{0}$, then the modified output (shortcut added) becomes
$$
\small
\mathcal{F}_B(\mathbf{x}) + \mathbf{x} = \mathbf{0} + \mathbf{x} = \mathbf{x}
$$
That's exactly the identity mapping! So in learning the easy zero mapping, we've created a block that corresponds to an identity mapping, which is a much harder task.
</div>


<h2>Chapter Goals:</h2>
<ul>
	<li>Understand the purpose of residual learning</li>
	<li>Learn the structure of a ResNet building block</li>
	<li>Create a function for a ResNet building block</li>
</ul>


Learn about the main building block of ResNet and how it works.

What you'll learn from this course

Image Processing

CNN

SqueezeNet

ResNet

ResNet Block

Chapter Goals:

A. Learning the identity