基于Mask Scoring R-CNN的高质量数据集快速自动标定方法

Rapid Automatic Labeling Method for High Quality Data Sets Based on Mask Scoring R-CNN

针对计算机视觉领域人工标定多目标数据集时间冗长的问题,提出一种基于Mask Scoring R-CNN的高质量数据集快速自动标定方法;首先,设计了高质量数据集快速自动标定架构,训练数据自动标定模型并搭建目标分类与标定系统;其次,在对比不同残差网络及引入迁移学习基础上,进一步研究了基于MaskIoU Head的多目标掩膜标定质量评价方法,完成基于Mask Scoring R-CNN的多目标高质量数据集快速自动标定方法设计;最后,以车辆数据为例进行数据集快速自动标定方法验证,实验结果表明,相较于Mask R-CNN和Faster R-CNN方法,Mask Scoring R-CNN方法具有目标数据分类效果好及掩膜分割精度高的优点,检测准确率达到93.4%,且标定速度相较于人工标定速度提升了95.77%。

Aiming at the problem of lengthy manual calibration of multi-target data sets in the field of computer vision,a rapid automatic labeling method for high quality data sets based on Mask Scoring R-CNN is proposed.Firstly,the rapid automatic labeling framework for high quality data sets is designed.Then,the automatic labeling model of multiple target data is trained,and the classification and labeling system for the target is built.Secondly,on the basis of the comparison of different residual networks and the introduction of transfer learning,the quality evaluation method of multiple targets mask labeling based on MaskIoU Head is further studied.Besides,the rapid automatic label method for multiple targets high quality data sets based on Mask Scoring R-CNN is implemented.Finally,the vehicle data is taken as an example,the data sets rapid automatic labeling method is verified.The experimental results show that compared with the Mask R-CNN and Faster R-CNN,the Mask Scoring R-CNN has the advantages of good effect of target data classification and high accuracy of mask segmentation,its detection accuracy reaches 93.4%,and the labeling speed of the method is 95.77%higher than that of manual labeling.