Lost-Found Item Net for Classification Based on Inception-Resnet
Yu‐xin Liu, Kai Fang, Enze Yang, Shitao Zhao, Shuoyan Liu, Ran He · 2022
A novel Lost-Found item net has been introduced in our paper. This net is help to classify the types of items in the clutter which collected in one train and waiting area of railway station. The objective of this net is contributing to a more efficient, comfortable ambiance for travelers. This paper develops and evaluates a rapid, low-cost system for item classification. Firstly, a high-quality lost-found items classification dataset is created, named Lost-Found Item Dataset (LFID). Secondly, our paper is introduced an improved architecture named inception-Resnet which is for the classification to recognize the item by extracting and learning some critical feature. In order to enhance the performance of our network, some significant data augmentation methods are applied to preprocess the item images. Through these novel technologies the important parameters of this model could be learned abundantly tend to generalization. Furtherly, a few experiments have been carried out and these performances have been discussed using this above dataset. The performance of our model has been well described and introduced. Finally, these results are compared with the most advanced methods available. The comparison with the popular methods shows the superiority of the novel model.