The effect of task on classification accuracy

摘要: Generating, grouping, and labeling free-sketch data is a difficult time-consuming task for both user study participants researchers. To simplify this process parties, we would like to have users draw isolated shapes instead of complete sketches that must be hand-labeled grouped, then use train our symbol recognizer. However, it an open question whether in isolation accurately reflect the way diagram. Furthermore, many simplest shape recognition algorithms were designed recognize gestures, not clear they will generalize freely-drawn shapes. answer these questions, perform experiments using three different recognizers measure effect collection on accuracy. We find trained only can classify freely-sketched as well same free-sketches. also show user-specific training examples significantly improve rates. Finally, introduce variant popular simple gesture algorithm recognizes highly-accurate but more complex recognizer explicitly recognition.