Algorithm Driven Development
Traditional Machine Vision, also referred to as classical computer vision, requires developing a custom algorithm every time that is rules based. These algorithms can be very sophisticated and require a highly trained and expensive engineer to program. Once that engineer finishes development and walks away from the project there is no way to improve the inspection. Even worse if the engineer tells you that he can only get to a subpar accuracy, there is nothing more you can do. So not only does traditional machine vision have a ceiling, but it can also only inspect limited and clearly defined tasks. Because of these fundamental issues and the fact that it can’t improve over time, traditional machine vision systems are decreasing in popularity compared to new deep learning systems.