Nuts and Bolts of Deep Learning for Computer Vision

Abstract

We humans are blessed with extraordinary powers of mind. These powers allow us to differentiate and distinguish, develop new skills, learn new arts, and make rational decisions. Our visual powers have no limits. We can recognize faces regardless of pose and background. We can distinguish objects like cars, dogs, tables, phones, and so on irrespective of the brand and type. We can recognize colors and shapes and distinguish clearly and easily between them. This power is developed periodically and systematically. In our young age, we continuously learn the attributes of objects and develop our knowledge. That information is kept safe in our memory. With time, these knowledge and learning improve. This is such an astonishing process that iteratively trains our eyes and minds. It is often argued that Deep Learning originated as a mechanism to mimic these extraordinary powers. In computer vision, deep learning is helping us to uncover the capabilities which can be used to help organizations use computer vision for productive purposes. Deep learning has evolved a lot and is still having a lot of scope for further progress.