I’m currently working on a research project focused on developing a Worker Assistance System. The project aims to automatically detect small damages on highly reflective and metallic surfaces. As part of my research, I’m exploring the feasibility of labeling objects based on their normal maps for the defect detection algorithm. The concept involves utilizing different normal maps corresponding to various defects present on objects.
Essentially, I want to know if it’s possible to label defects by analyzing the information encoded in the normal maps? I’ve already experimented with the perception package, but unfortunately, it didn’t yield the expected results.
Has anyone here had similar experiences or could provide recommendations on how to approach this challenge?
Your insights would be greatly appreciated. Thank you!