While evaluating models I have encountered quite a few samples that I am having trouble with. One example is below.
Obviously everyone is equally impacted by the presence of images like this but out of curiosity are annotations like this incorrect or is this correct (based off of the spacecraft model)? I need to investigate quantitatively but with the naked eye there does not seem to be any signal that could facilitate the correct prediction of the true box (for the image above).