Classification Algorithm for Multiple Image Formats Suorting Complex Neural Network Models

Abstract

In recent years, with the continuous exploration of NNs and the upgrading of mobile phone hardware, more and more researches focus on how to design effective models to suort terminal task reasoning. IC is a classic task in many studies. It can be used in many fields such as image search, face recognition, and medical imaging. It has great practical significance in reality. The main purpose of this paper is to carry out optimization research on the classification algorithm of multiple image formats based on the suort of complex NN models (NNM). This paper proposes improved methods for problem transformation and model adaptation, and studies the problems of blurred scenes and difficult method selection in multi-format image classification (IC). In the direction of problem transformation, a basic classification model and a classification model combined with advanced strategies are designed., in the direction of model adaptation, transform multi-format output and use transfer learning for training. Finally, it is found that the problem transformation method can obtain better classification results in small data sets, but the classification time is several times as long as the model adaptation method. The model adaptation method can save a lot of time and is easier to promote.