Effect of attention and triplet loss on chart classification: a study on noisy charts and confusing chart pairs

Charts are powerful tools for visualizing and comparing data. With the increase in presence of various chart types scientific documents electronic media, development an automatic classification system is becoming important task. Existing studies on fail to address noise charts confusing class pairs. Motivated by above observations, this paper, we propose attention triplet loss based deep CNN framework issues. From experimental results over four datasets, it evident that proposed can effectively handle samples outperforms its counterparts.