I'm training a pretrained mask r cnn network (maskrcnn_resnet50_fpn) in pytorch with my own data. Goal is, that the network can detect one object class from images.
I took my pictures and now I'm in the process of labeling them. In the pictures I took so far, there is only one object per picture. Now I was wondering if the training of the network would profit from having the same object multiple times in one picture. So in my case I want to detect a special type of gear and I have multiple gears of the special type laying around. I photographed them one by one from different perspectives and angles, but not multiple gears of the special type in one picture. Should I do so?
