How to evaluate a Deep Learning solution
If I were to work with a machine vision engineer or a plant manager familiar with machine vision as an enabling technology in automation, and try to explain to them how they could approach certain applications with deep learning, I would start by advising them to analyze the application and understand what value they intend to get out of the solution. What are the needs of the application? What are the needs of the process related to this individual application? And then apply those needs, those specifications if you will, apply them to the selection of a final product or a final system to be implemented for that application. That system may be deep learning and with the right application, it should be deep learning, on the other hand, it may be traditional machine vision or rule-based machine vision. And we want the customer, I think everybody in the industry wants the customer to implement the right solution, the most efficient solution for their individual application. I think that we will find a wide range of applications suitable to both, and deep learning is only enhancing that group of applications that can be solved on the plant floor.