Edge detection is essential in every aspect of computer vision from vehicle number plate recognition to object detection in images or video. Preprocessing stage of many image processing tasks usually associated with Artificial Intelligence are relying on separating multiple objects from each other before accessing any other information. Over the years, researchers have used Sobel, Prewet or Roberts filter and then relying more on robust Canny method. Today, there have been many improvements on the earlier approaches, and now very much rely on Convolutional Neural Networks to assist in determining effective edges that would assist immensely in object detection. From that perspective, it is quite evident that effective edge detection is all about eventual object detection. With this notion in mind, it is easy to see what methods work and what methods would not achieve the goals. Deep Learning (DL) approaches have been gaining popularity over the years. Do DL algorithms outperform the conventional edge detection algorithms? If they do, is it time for us to forget about the conventional approaches and resort to the new state-of-the art? Are they transparent, when performing poorly? How reliable are they? Do they perform consistently when unknown data are presented? This article will analyze the existing and emerging edge detection methods with a view to determine their usability and limitations in computer vision applications that would undoubtedly advance the field of image processing and computer vision.