Simulation of Multi Label Classification of Natural Scene Images with Improved GoogLeNet
In order to improve the accuracy of multi-label classification for natural scene images,this paper pres-ented a multi-label classification method for natural scene images based on improved GoogLeNet.At first,a boot im-age was constructed by the weighted least squares filter in the adaptive guided filter algorithm,and then the existing filter was modified.Moreover,the gradient offset was added to carry out the pixel adaptation and highlight the edge of the natural scene image,thus realizing the detail enhancement.Based on the characteristics of natural scene image,a multi-label category library of natural scene images was constructed to improve the structure of GoogLeNet network.After that,all images were input into different GoogLeNet models.Finally,the multi-label classification for natural scene images was achieved.The experimental results show that the proposed method can effectively improve the accu-racy and classification efficiency of multi-label classification result of natural scene image.
Natural scene imageMulti-label classificationImage enhancement
万方数据
维普
