Improved Fashion Classification Method Base on GooLeNet

Traditional Image Classification algorithms have problems of high recognition error rate and low efficiency in fashion classification. So, the research proposes an improved model based on GooLeNet model.At first, we created a new inception network module which called inception-improved module.Compared with original inception module,Inception-improved module reduced the computational cost and improved the efficiency of the network.Secondly, we used Inception-improved modules to build the new model which named GooLeNet-improved and trained the model with Fashion image dataset. The experimental results show that the improved model GooLeNet-improved can obviously reduce the error rate of the fashion classification and improve the computational efficiency. The GooleNet-improved model achieved the accuracy of 87.1% on fashion image dataset, and the accuracy is respectively 0.8% higher than the original GooLeNet model.