Here at team homer we start the new year by spending hours labelling cups, cutlery and long-expired food items in pictures. Because we know how to have fun.
For those who aren’t familiar, labelling simply means to mark the region in an image where a specific object is. We do this for many pictures and then feed the labelled images into a neural network. This way, the network learns how to recognize the trained objects in new pictures even with perspectives not seen before.
We labelled more than 300 pictures of our many test objects, so Lisa and TIAGo should now be better equipped for recognizing and grabbing the right objects in the coming tests.