I. Documentation names and authorsconvolutional neural Networks at Constrained time COST,CVPR
two. Reading timeJune 30, 2015
Three. Purpose of the documentThe author hopes to improve the accuracy of CNN by modifying the model depth and the parameters of the convolution template, while maintaining the computational complexity. Through a lot of experiments, the author finds the importance of different parameters in the network structure, and obtains the competitive effect on the ImageNet2012 data set.
Four. Contribution points of the literatureThe author's contribution is mainly to illustrate the effect of different parameters on the accuracy rate through various comparative experiments. The contribution point of theory is comparatively few. Through experiments, the author obtains the following two phenomena about depth: 1. Depth is the first factor that affects the accuracy rate; 2. Although depth is very important, the accuracy rate will decrease if the depth is too deep.
Five. The database usedImageNet2012 Data Set
Six. Experimental resultsThe main part is the comparison of all kinds of modified models, and the comparison with the mainstream algorithm mainly focuses on the accuracy rate is not low at the same time, the computational complexity is also relatively low.
Copyright, welcome reprint, reproduced Please indicate the source, thank you
Copyright NOTICE: This article for Bo Master original article, without Bo Master permission not reproduced.
convolutional neural Networks at Constrained time Cost (intensive reading)
