摘要
随着计算机视觉技术的飞速发展,传统检测行业中图片的自动识别与分类功能得以实现。通过深入分析多种卷积神经网络模型的性能特点及汽车产品检测的实际工作需求,选取了VGGNet模型进行结构搭建与模型训练。经过多次细致的参数调整与优化,最终版模型能够精准地识别出企业备案照片中车辆的朝向。模型的Accuracy(准确率)与平衡F分数(F1-Score)均高达0.95,这一成果对于推动科技创新技术在汽车检测行业的具体应用具有重要意义,有助于加快实现行业的“科技赋能、创新发展”。
With the rapid development of computer vision technology,automatic image recognition and classification in the traditional inspection industry has become possible.By analyzing the performance characteristics of various convolutional neural network models and the work requirements of automotive product testing,this paper constructs and trains a VGGNet model.By adjusting parameters multiple times,the final version of the model is able to accurately identify the vehicle orientation in photos in the record.The accuracy and F1-Score of the model are both as high as 0.95.This work is conducive to promoting the specific application of scientific and innovative technology in the automotive testing industry,and accelerating the realization of“technology empowerment and innovative development”in the industry.
出处
《商用汽车》
2024年第2期82-87,共6页
Commercial Vehicle
关键词
图像分类
卷积神经网络
汽车产品检测
Image classification
Convolutional neural networks
Automotive product testing
