The broken wire model was trained with 209 "augmented" images of broken wires agains 90 images of things you deemed not to be broken wires and many of them had other issues like missing pedals and stuff. I need a way to systematically test the effectiveness of this training method and I think a good start is to train it against other augmented images.
The broken wire model was trained with 209 "augmented" images of broken wires agains 90 images of things you deemed not to be broken wires and many of them had other issues like missing pedals and stuff. I need a way to systematically test the effectiveness of this training method and I think a good start is to train it against other augmented images.