基于视觉大模型的哑资源异常识别方法
Dumb Resource Anomaly Detection Method based on Visual Foundation Model
尹俊 1刘贤松 1钟晨杰 1陶颖1
作者信息
- 1. 中国联通智网创新中心,上海 200050
- 折叠
摘要
文章分析了哑资源巡检以及人工智能技术应用于运营商智能巡检的现状,并介绍了基于vision-transformer技术的Swin大模型.基于该模型的预训练版本,结合标注好的图像数据,完成了针对运营商哑资源分类检测的模型训练并评估了模型效果.展望了该模型的优化方向以及大模型应用于运营商场景的前景.
Abstract
The artical analyzes the current situation of dumb resource patrol and the application of AI technology to intelligent patrol of operators,and introduces the Swin foundation model based on vision-transformer technology.Based on the pre-training version of the model,combined with the labeled image data,the model training for the operator's dumb resource classification is completed and the model effect is evaluated.The optimization direction of the model and the prospect of the application of the large model to the operator scenario are prospected.
关键词
大模型/人工智能/机器视觉/哑资源/智能巡检Key words
foundation model/artificial intelligence/machine vision/dumb resource/intelligent patrol引用本文复制引用
出版年
2024
NETL
NSTL
万方数据