I have theorical question that I couldnt decide how to approach. I have tons of grayscaled shape pictures and my goal is seperate these images to good printed and bad printed. For this, I look at roughness of images because some of those is corrupted while printing (e.g I expected triangle shape but I saw some corruption in its edges. So its bad printed). Should I train my model to give each of them spesific label names (like good_printed_triangle, bad_printed_triangle, good_printed_square etc.) or just good and bad is ok? I was thinking about second option but how can CNN find correlation between different shapes, it makes me confuse. I am open for all ideas and thank you in advance.
In general in this situation you should be training to just the specific problem you are trying to solve - which from the sounds of it is just whether edges of any kind on the shape are blurry or cleanly printed.
Unless you specifically need the shape labelled as a square triangle etc you should drop it as a label as will just add unnecessary complexity and overhead to the model.