中国计量大学学报2018，Vol.29Issue(4)：452-456,5.DOI:10.3969/j.issn.2096-2835.2018.04.017
一种改进的残差网络宫颈癌细胞图像识别方法
Establishment of an improved residual network method for recognition of cervical cancer cell image
摘要
Abstract
An improved residual network (ResNet) algorithm was proposed to reduce the false negative rate of cervical cancer cell image recognition in this paper.The ResNet algorithm increased the weight of the cross entropy cost function to establish the weight matrix according to the cervical cells of different degrees of lesion.The output of the false negative category was weighted to reduce false negative judgment.The experimental results showed that the algorithm classification is stable for different cervical cell image data sets.Compared with the traditional image classification algorithms, the improved cross entropy cost function algorithm can effectively reduce the false negative rate of cervical cancer cell image recognition.关键词
残差网络/图像识别/交叉熵代价函数/宫颈癌细胞/假阴性率Key words
residual network/image recognition/cross entropy cost function algorithm/cervical cancer cell/false negative分类
信息技术与安全科学引用本文复制引用
谢欣,夏哲雷..一种改进的残差网络宫颈癌细胞图像识别方法[J].中国计量大学学报,2018,29(4):452-456,5.基金项目
浙江省自然科学基金项目(No.LY12F01011) （No.LY12F01011）
