Determining a good deep learning project versus a traditional machine vision approach not only involves the application analysis but it involves the end result that has to be achieved. We can follow some rules when we look into these applications, there are some guidelines if you will. The deep learning application is almost always very well indicated when there is subjective material, subjective features, whether it be defects or anomalies that have to be detected, or features that have to be classified where a human does best at classifying those, but really can’t tell you exactly why they were classifying it that way. Rule-based machine vision works best when it is doing discrete analysis such as measurement, detection of features that are known and quantifiable, and reporting on features based on their size, geometry, and other characteristics. There of course are gray areas, but those are some of the basic deep learning examples of selecting the right application for the right technology.
View all Videos
How to determine a good Deep Learning and Machine Vision project?