Draft drawings

This study used boots, aircraft, cells, pliers, and guitars from 2D shapes included in the literature as data sets to test modeling success. These 2D shapes, which are mostly not publicly available data sets, form the target curves of IP. In this study, hand drawings with a curved structure were used in modeling, where the success of fitting precision could be better determined. Other reasons for preferring hand drawings as a data set are that these images provide data diversity by covering noisy examples in the real world, can improve the model’s generalization ability, and can reduce the overfitting problem. Draft drawings are provided in Supplementary files.

本研究选取了文献中包含的靴子、飞机、细胞、钳子和吉他的二维形状作为数据集，以检验建模的成功度。这些二维形状，大部分并非公开的数据集，构成了IP的目标曲线。在本研究中，采用了具有曲线结构的手绘图像进行建模，从而更精确地确定了拟合精度。偏好手绘图像作为数据集的其他原因还包括，这些图像通过覆盖现实世界中的噪声示例，提供了数据多样性，能够提升模型泛化能力，并有助于减轻过拟合问题。草稿图可在补充文件中找到。