One of the reasons Machine Learning is such an interesting field is it allows us to apply computing logic to areas that previously were untouchable. While computers are extremely effective with arrays and integers, they have traditionally been less apt at dealing with emergent properties. For example, you cannot look at just one pixel on a screen and know the image is a dog. You have to synthesize lots of data points.
In the past decade, computer scientists were able to bridge this divide by creating Computer Vision models— specifically Convolutional Neural Networks (CNNs). Today, I’m going to show how to apply them to image classification.
Classification of real world data is very useful for integrating machine learning technology into more typical software systems. If you’re in e-commerce, you may use this information to automatically categorize a new product. If you’re in medicine, you may use this to determine if an X-Ray or MRI looks similar to previous images that required surgery. Finally, if you’re in a vehicle and looking to drive safely, image classification is a key part of object detection and collision avoidance.