Research On Furniture Image Style Classification Model Based On Deep Learning

In the field of e-commerce applications,furniture image recognition technology has great scientific research potential and commercial value.With the development of computational aesthetics research,scholars have begun to pay attention to the study of image semantic information such as style.Among them,the furniture image style recognition technology,which studies furniture recognition from the perspective of furniture style information,is an important supplement to the current furniture image recognition work.At present,there are few researches on furniture image style recognition,and the application scenarios are different.The main difficulties and problems are as follows:1.Most of the existing furniture image recognition technologies carry out research from the perspective of the category of furniture itself(such as tables and chairs),while ignoring the important attribute information of furniture style.There is little research work on furniture style classification,and there is no public database of furniture image styles.2.The definition of style features is ambiguous and highly subjective,and it is difficult to describe qualitatively.There is no corresponding measurement standard for different furniture style features,which is not conducive to feature extraction and analysis.3.The style attribute information of furniture images is very complex,and it is very difficult to identify furniture images according to different styles.At present,there are few recognition methods for furniture style recognition and the recognition accuracy is low.4.The application of existing recognition models varies greatly,and there is no recognition model specifically for the style of furniture images.Aiming at the above problems,the main research work and innovation of this paper are:1.Build a database of furniture image styles.For furniture product categories with clear styles in furniture images,collect furniture images published on e-commerce websites to establish a database,and do basic work for subsequent research on furniture image style recognition.2.Measure and analyze furniture images of different styles.Qualitative description and quantitative analysis of different styles of furniture from the perspective of image features.3.Extraction and recognition of furniture image style features.Based on the style feature measurement work,a furniture style recognition method based on Gram transform and feature fusion is proposed.4.On the basis of the proposed furniture style recognition method,use the convolutional neural network technology to design the FSICM model(Furniture Style Image Classification Model)specially used for furniture image style recognition,and adjust and optimize the parameters of the model.The FSICM model is trained and verified through the established data set,and the recognition accuracy of the model for the training set and test set is 98.40%and 92.80%,respectively.Compared with other classic convolutional neural network models,the recognition accuracy of the FSICM model has been greatly improved,the model fitting effect is good,and the classification accuracy is high,which fully proves the effectiveness of the built model in the task of furniture image style recognition.