基于目标检测的图形用户界面控件识别方法

林灏昶, 秦云川, 蔡宇辉, 李肯立, 唐卓

Graphical user interface widget extraction based on object detection

Haochang Lin, Yunchuan Qin, Yuhui Cai, Kenli Li, Zhuo Tang

表2 在不同数据集上目标检测模型的测试结果

Table 2 Experimental results of object detection by different algorithms on different datasets

数据量 模型 mAP@0.5 mAP@0.5∶0.95 2000 SSD 92.12 73.21 YOLOv3 93.27 76.41 YOLOv4 92.09 78.92 3000 SSD 92.67 74.53 YOLOv3 92.67 76.05 YOLOv4 94.95 80.23 4000 SSD 92.76 74.66 YOLOv3 93.08 77.13 YOLOv4 94.03 77.77 5000 SSD 93.04 74.62 YOLOv3 95.84 80.03 YOLOv4 97.79 81.44