modify vision rubrics in main cases, add overall visualization goals to every cases, and remove rubrics that cannot be rated by comparing GT and result images

first commit