Segmented Inscription Recognition Index for Measuring the Inscription Quality Extracted From 3-D Data
Beom-Chae Jeong, Ye-Chan Choi, Sheriff Murtala, Kang-Sun Choi · IEIE Transactions on Smart Processing and Computing · 2020
This paper proposes a novel measuring method for evaluating the inscriptions extracted from the three-dimensional scanning data of monuments based on the ground truth engravings. The inscription prediction affects inscription recognition differently regarding the prediction region. That is, regional importance in inscription recognition varies. Based on the observation, This paper proposes to divide the 2-class labeled ground truth into four areas according to the regional importance. For each classified area, the true positive, false negative, false positive, and true negative were weighted differently and used to determine a modified confusion matrix. After configuring the confusion matrix, the F1 score was calculated. The experiments showed that the proposed measure was more similar to the subjective inscription recognition by people than the mean opinion score.