Machine Learning in Computer Vision

Introduction

Machine Learning is the most cited word in past decades for computer vision research. The availability of enormous data and the demand for analyzing data makes Machine Learning a must-have option for each data scientist. Nowadays, it is hard to find a data analysis that has not done using Machine Learning. It is pretty extraordinary that Machine Learning opens up a new era for high-performance data analysis and sometimes beats the human brain (Steinberg, 2017). Similar to other research areas, Computer vision benefits from Machine Learning, so many computer vision researchers use Machine Learning for their research. In computer vision, the problem domain varies a lot. Image segmentation, classification, recognition, feature extractions are some problem domain that are most frequently used. Before the blooming of Machine learning, other techniques- filtering, wavelength analysis, Histogram processing, Fourier transformation, morphological algorithms were used for specific task. Nonetheless, images vary a lot and each of those methods has their limitations to success while doing specific job.

Machine Learning is not a new concept. Donal Hebb introduced the concept in a book titled” The Organization of Behavior.” In 1949. In this book, the author introduced a model based on human brain cell interaction. Hebb wrote, “When one cell repeatedly assists in firing another, the axon of the first cell develops synaptic knobs (or enlarges them if they already exist) in contact with the soma of the second cell.” (Keith, 2019). However, neurophysiologist McCulloch and Mathematician Walter Pitts invented the concept of the human brain neuron in 1943. The famous Alan Turing used the same concept for creating the “Turing Test.” Frank Rosenblatt designed the first artificial neural network- “Perceptron” for pattern and shaped recognition in 1958. Unlike the other breakthrough technology, Machine Learning was dormant in the decades of 1980 and 1990. Tech giants like IBM and research laboratories- AT&T Bell gave rebirth to Machine Learning at the beginning of the 21^st century. In the 21st century, the most significant computer vision breakthrough using Machine Learning was Alexnet (Krizhevsky, 2012) and GoogleBrain (2012). The Alexnet was the winner of the ImageNet competition. It is mention-worthy that ImageNet is a competition for image classification with a vast database of images. After the competition, many computer vision researchers used it as an opportunity to develop models for various computer vision applications using Machine Learning. In this chapter, the main focus will be on Machine learning algorithms and their various implementation on computer vision. Besides, relevant comparison with pre-existing methods will be discussed in a brief.