Cultural relic classification method based on lightweight deep learning network

Abstract

The invention provides a cultural relic classification method based on a lightweight deep learning network, and the method comprises the steps: firstly constructing a data set for the classification of cultural relic images, and then dividing an original cultural relic image data set into a plurality of relatively independent small data sets through employing a coarse-grained clustering neural network, for each small data set, using a lightweight neural network classifier designed by the invention to classify data, and finally, using an energy prediction model to measure whether the CNN model can be deployed on a device with limited resources. According to the method, the model structure can be simplified, the training speed can be increased, the reusability is higher, the precision and the calculated amount can be better balanced, and the method is suitable for being deployed on equipment with limited resources.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence and relates to the field of deep learning, in particular to a method for classifying images of cultural relics, which aims to provide museums with high-precision and stable classification of cultural relics based on images. Background technique [0002] Efficient automatic classification of cultural relic images is one of the key technologies of cultural relic image big data. At present, the websites of major museums in China mainly use keywords to retrieve images, and first of all, it is necessary to manually label images of cultural relics. With the sharp increase in the number of images of cultural relics, this manual method will bring great costs. Therefore, the use of artificial intelligence technology to intelligently distinguish the types of cultural relics and classify cultural relics into corresponding entries can greatly improve the efficiency of cultural relic classific...

AI Technical Summary

Problems solved by technology

Howard et al. (Howard AG, Zhu M, Chen B, Kalenichenko D, Wang W, Weyand T, Andreetto M, Adam H. Mobilenets: efficient convolutional neural networks for mobile vision applications[J].arXiv,2017:1704.04861.) proposed MobileNet V1 depth separable convolution, its essence is to sparsely express redundant information, but its structure is too simple, resulting in low cost performance of the network

Images