帧(网络)
计算机科学
人工智能
建筑
人工神经网络
计算机视觉
工程制图
模式识别(心理学)
工程类
视觉艺术
艺术
电信
作者
Alexey Mihajlovich Kashevnik,Ammar Ali,Alexander Mayatin
标识
DOI:10.1109/smartindustrycon57312.2023.10110731
摘要
The paper proposes a method for frame detection in engineering drawings documents. The method is based on the modern AI techniques. We proposed to use YOLOv5 as framework for neural network architecture development for frame detection. We presented a neural network architecture, introduce loss functions we used and show the experiments. We got good results showed that our accuracy is about 99% for frame detection. Sure that we used similar drawings for training and testing dataset. We showed examples in the paper how is the frame detection worked for several examples of drawings.
科研通智能强力驱动
Strongly Powered by AbleSci AI