摘要
通过对YOLOX-S模型引入可变形卷积神经网络和焦点损失函数(Focal loss),解决原YOLOX-S模型车窗识别准确率较低的问题.首先,通过在YO LOX-S模型的主干特征提取网络中引入可变形卷积神经网络,对卷积核中的各采样点引入偏移量,以便在原始图像中提取到更具有表征的信息,从而提高车窗识别的精准度;其次,使用Focal loss替代原模型中的二元交叉熵损失函数,Focal loss能缓解正负样本不平衡对训练的影响,其在训练过程中更关注难样本,从而提高了模型对车窗目标的识别性能;最后,为验证改进算法的性能,实验收集并标注15627张图片进行训练和验证.实验结果表明,改进后的车窗识别算法的平均目标精度提高了3.88%.
We solved the problem of low accuracy in car window recognition of the original YOLOX-S model by introducing deformable convolut ional neural networks and Focal loss function(Focal loss)to the YOLOX-S model.Firstly,by introducing deformable convolutional neural networks into the backbone feature extract ion network of the YOLOX-S model,offsets were introduced for each sampling point in the convolutional kernel to facilitate the extraction of more representati ve information from the original image,thereby improving the accuracy of car window recognition.Secondly,using Focal loss instead of binary cross entropy loss function in t he original model,Focal loss could alleviate the impact of imbalance between positive and negative samples on training,and it paid more attention to difficult samples during the training process,thereby improving the recognition performance of the model for car window targets.Finally,in order to verify the performance of the improved algorit hm,15627 images were collected and annotated for training and validation in the experiment.The experimental results show that the average target accuracy of the improved car wi ndow recognition algorithm increases by 3.88%.
作者
黄键
徐伟峰
苏攀
王洪涛
李真真
HUANG Jian;XU Weifeng;SU Pan;WANG Hongtao;LI Zhenzhen(Department of Computer,North China Electric Power University(Baoding),Baoding 071003,Hebei Province,China;Hebei Key Laboratory of Knowledge Computing for Energy&Power,Baoding 071003,Hebei Province,China)
出处
《吉林大学学报（理学版）》
CAS
北大核心
2023年第4期875-882,共8页
Journal of Jilin University:Science Edition
基金
国家自然科学基金(批准号:61802124)
全国高等院校计算机基础教育研究会项目(批准号:2019-AFCEC-125)
中央高校基本科研业务费专项基金(批准号:2021MS089).
