Resource download address : https://download.csdn.net/download/sheziqiong/87904742
Resource download address : https://download.csdn.net/download/sheziqiong/87904742
1. Background and significance of topic selection
1. Subject research Background
With the rapid development of mobile terminal technology and network technology, people can use mobile clients to surf the Internet, obtain information and services from the Internet anytime and anywhere, and solve problems in all aspects of life, such as food, clothing, housing, transportation, and social and physical health management. All kinds of mobile application services emerge as the times require.
At the same time, with the development of the economy and the improvement of living standards, people's consumption of clothing has increased sharply. In the apparel sales industry, according to the "China Apparel E-Commerce Industry Development Report" [1] released by the China Electronic Commerce Center, the transaction volume of my country's apparel e-commerce in 2015-2016 was 3.7 trillion yuan, compared with 2014 The 2.9 trillion yuan in 2019 has increased by 25.4% year-on-year. At the beginning of 2022, my country's textile and apparel exports will increase by 8.9% year-on-year, and the export of most categories of textile products still maintains a relatively high growth rate [2]. With the development of the fashion industry and the promotion of niche culture, people's requirements for clothing are no longer limited to enough to wear, but the style, color, material, etc. of clothing, the diversity of clothing and the demand for personalized consumption are also gradually increasing. . The research on clothing images has been continuously developed along with the rise of mobile shopping platforms. At present, in the field of visual fashion analysis of clothing e-commerce, the research mainly focuses on clothing image classification, recognition, attribute prediction, retrieval, collocation, clothing recommendation and intelligent fitting. At the same time, Baidu, Amazon, Google and other technology companies have launched applications and products with image search functions. In clothing image search and applications, Alibaba’s various apps have also launched the “search image by image” function[3] . Generally speaking, clothing image recognition and management technology is constantly improving, and the threshold for use is also constantly lowering.
2. Significance of the subject research
The continuous increase in the number of clothing consumption means that the average number of clothing purchased by a single person is also increasing. Online and offline stores and second-hand transactions rarely support customization. There is a gap between the consumer demand for clothing products and product design and supply [4]. People's individual needs are also facing the problem that they cannot be met in time. It is also difficult to improve. It can be seen that the development of a mobile terminal application system that supports personalized clothing customization can not only meet customers' convenient and personalized shopping needs anytime and anywhere, but also meet the business needs of enterprises, optimize enterprise management methods, and improve enterprise work efficiency. Thereby promoting the further development of the clothing customization industry.
On this basis, the mobile terminal system can comprehensively use image recognition, image classification, retrieval and recommendation and other relatively mature technologies to form information management of clothing, and realize data import and classification management of images of purchased clothing taken and uploaded by customers. , visualize the results, search according to the customer's description of the target clothing, and automatically recommend according to the degree of conformity. If the customer does not find a satisfactory clothing, the description will be converted into a purchase post, which will largely solve the current clothing problems faced by customers. management issues. The research direction of this topic is the design and implementation of clothing identification and management system, which is aimed at customers who need clothing management and personalization as well as various clothing suppliers. The final presentation is an Android-based APP, which is in line with the current clothing Pain points of information management and sales market.
2. Research status and development trends at home and abroad
1) Regarding clothing recognition and management,
a complete image recognition task process is mainly divided into three steps: ① The image segmentation stage is used to complete the separation of the target object and the background image for a given image ; ② The feature extraction stage is used to extract the features that can represent the target from the segmented image; ③ The image classification stage that classifies the target according to the extracted features. With the gradual maturity of image processing technology and the rapid development of computer vision technology, image recognition technology has been perfected and used more and more widely [5].
As the most popular method in machine learning, deep learning originated from the artificial neuron perception model and was popular in the back-propagation (back-propagation) in the 1980s [6]. In 2012, Krizhevsky et al. proposed AlexNet deep convolution Convolution Neural Network (CNN) [7] achieved the best results at that time in the ImageNet image recognition competition with a significant advantage and surpassed the accuracy of human recognition. In August 2017, OpenCV 3.3 was officially released, bringing a highly improved "Deep Neural Network" (DNN) module, which supports a variety of deep learning frameworks, including Caffe, TensorFlow, and Torch/PyTorch, for image realization through deep learning. provide great support for the identification and management of With the promotion of big data and the improvement of the deep learning framework in recent years, many domestic and foreign scholars have begun to apply deep learning technology to clothing-related research. The article Where-to-Buy-It ( In WTBI) [8], it is proposed to formalize the street-to-shop clothing recognition and retrieval task as a cross-domain product similarity learning problem, and on this basis, a similarity model for specific categories is designed. Network parameter learning method for degree calculation. Chinese Taiwanese scholar Lin[9] obtained hash-like image feature representation by fine-tuning the pre-trained model, and then performed image retrieval through hierarchical deep search. After 2017, with Baidu EasyDL as the mainstay in China, a number of platforms that provide commercial AI model training and services for users with zero AI algorithm foundation or pursuit of high-efficiency development have been gradually established and popularized. The accuracy of various models after training on the platform high and rising.
Clothing information management, relying on clothing identification, means that users can completely and accurately grasp the clothing information they own, and realize the reorganization and reuse of clothing information, so as to solve practical problems in life. According to the prototypes designed in some literatures and the products on the market, currently, solutions to this requirement can be divided into three categories: (1) general information recording tools such as Excel, Evernote, etc. (2) pure software forms Clothing management apps such as "Today's OOTD", "Stylebook" and so on; (3) Clothing management systems that combine software and hardware. For "Smart Closet", "Smart Wardrobe". These have realized the information management of clothing to a certain extent, but none of them support personalized custom clothing.
2) About clothing customization
Clothing customization can be divided into five stages in chronological order: the original manual stage, the shrinking era of manual customization (the rise of ready-made garments), the elimination era of manual customization (the popularity of large-scale differentiated ready-made garments), and the new customization The embryonic period (the dilemma of ready-made clothing), the outbreak period of new customization (the large-scale popularization of tailor-made customization). Neither ready-to-wear nor custom-made clothing can completely replace each other at present, but the custom-made clothing industry is in a spiraling stage[10].
The current mainstream clothing customization systems in foreign countries are mainly divided into two categories: one is the management system of high-end fashion clothing customization studios. This kind of studio has high requirements for craftsmanship, high prices, and a small audience. Purchase platform; another type of clothing customization system is applied in large clothing enterprises, which solves the mass customization production method of one person and one cut, which mainly relies on 3D human body scanning technology to store the customer's body data in the database of the management system , so that customers can have a virtual try-on, and then quickly pattern-making production through clothing CAD, CAM and automatic cutting machine technologies [9]. Satisfactory services are often not available. The current domestic suppliers for the clothing customization market are mainly concentrated in small and medium-sized enterprises and individual handicraft households. The scale is relatively small, and they cannot reach mass production. The clothing design education in China has the characteristics of single style and insufficient personalization. The cost of directly purchasing the "single-person single-cutting" system of large foreign companies is relatively high, which may be more expensive than self-development. Therefore, it is necessary to develop domestically more suitable for domestic customization. demanded sales system. At present, the common domestic customization needs are "customization by posting pictures", which is only for the customization of patterns, and the layout cannot jump out of the basic clothing styles. In addition, film and television dramas and opera stage costumes often need to be customized to better fit Character images, similar cheongsam circles and Hanfu circles have also set off a wave of customization [11].
To sum up, the current apparel identification and management APP has a good application prospect and development trend in the apparel industry. This type of APP can effectively solve the problems of unbalanced supply and inconsistent information management faced by the apparel industry. The mobile client comprehensively uses technologies such as image recognition, image classification and retrieval recommendation to realize the data import and classification of clothing images, visualize the results, search according to the description, and automatically recommend according to the degree of conformity, which can not only effectively carry out clothing information It also makes the rules of clothing classification clearer, which is of great help to the regulation and promotion of clothing wholesale, sales, design and other aspects. With the change of technology, the future clothing recognition technology can become a professional dressing consultant, and quickly get multiple sets of suitable matching combinations.
3. The main content of the research and the key problems to be solved
This graduation project aims to realize efficient clothing selection and classification management on the basis of informatization of various clothing. The target user group is people who have individualized needs for clothing. The system is developed with an Android client and a web management terminal, which have the advantages of being portable and interactive at any time. By collecting the picture information and personalized needs uploaded by users, a record of existing clothing and customization needs is formed. Based on these data, the development of data visualization system and management system can achieve the effect of recommendation and management.
More specifically, the main research contents and key issues to be addressed are as follows.
1. Research objectives and main content
1) Research and analysis of user needs
Apparel identification and management APP mainly has three types of users: First, customers who need clothing management and customization can count the existing clothing of customers for efficient classification and memory to avoid duplication The phenomenon of purchasing can also recommend suppliers for the individual needs of customers, so as to obtain satisfactory clothing; the second is all kinds of clothing suppliers, before producing the next batch of clothing, the suppliers collect all kinds of custom It is necessary to plan the next step of production, and at the same time expand the sales target to effectively solve the problem of unsalable products; the third is the system administrator, who manages the clothing information uploaded by customers and suppliers, the communication message information between the two, and the user account password information.
2) System function design
On the basis of a comprehensive demand analysis of clothing identification and management APP, in-depth analysis and design of the overall function of the system, database design and front-end interface effects, so that the final developed APP can achieve the following five important Functional modules: ①Data collection: shooting and sorting all kinds of clothing; ②Data import and preprocessing: develop data import and processing programs, and process them into predetermined formats and sizes according to parameters; ③Apparel learning and classification: use depth Learn the API and introduce the pre-training model to learn the imported clothing pictures and analyze their categories; ④ Clothing search: Search according to the customer's description of the target clothing, and automatically recommend according to the degree of conformity. If the customer does not pick a satisfactory clothing, it will be The description is turned into a buying post; ⑤Customized recommendation: Recommend customers' customized needs to suppliers who have uploaded related types of self-made clothing, so as to facilitate their point-to-point sales
3) Realization of the system
The realization of clothing identification and management APP is mainly divided into two aspects: system architecture realization and function realization. Among them, the system architecture adopts the front-end and back-end separation framework Android+Django+MySQL, and uses the OpenCV library; the function realization is mainly reflected in the five customers of "registration and login", "clothing upload", "clothing classification management", "clothing search" and "customized recommendation" The normal use of the end function module. System testing mainly tests whether the system functions are perfect and the performance meets the requirements, so as to ensure smooth and stable operation.
4) Summary and comprehensive analysis Analyze and
summarize the construction of apparel identification and management APP, comprehensively analyze the advantages and disadvantages of the system, and look forward to the application effect and development prospects of the system, and put forward the shortcomings of the system.
2. Key issues to be solved
1) How to reasonably preprocess and store the photographed clothing pictures?
2) How to achieve accurate results of clothing keyword retrieval matching?
3) How to assist users to realize clothing customization?
a) How to standardize the personalized customization needs of customers so that customers can communicate effectively with suppliers?
b) How to realize the quick selection of customers who can accept orders by suppliers?
4. Research methods and technical routes
1. Research methods
This topic mainly adopts three research methods.
1) Literature research method
Use the literature research method. By reading, translating and sorting out relevant literature and materials, we can have a clear understanding of the current development of identification and management systems at home and abroad and the technologies adopted. Through the case study of various mature technologies and software systems, the development of each functional module is carried out for reference.
2) Questionnaire survey method
Use the questionnaire survey method. By obtaining first-hand real information, as the analysis basis for user clothing management and personalized customization, the objects of the questionnaire survey are mainly college students and social workers.
3) Empirical research method
Experimental method is used. After understanding the main content of the research and formulating the technical route, continuous quantitative analysis and experiments are an indispensable part of the subject research. In this topic, an APP with Android as the carrier and a web management terminal are developed. The project is divided into three types of users: customers, suppliers, and administrators. Set up multiple users with different needs for comprehensive testing and use. View Whether the functional modules of the digitization and management of clothing information developed can be used normally.
2. Technical route
The overall technical route for the design and implementation of the apparel identification and management system is shown in Figure 1. In the first stage of this project, the problems to be solved are confirmed on the basis of analyzing the background and significance; in the second stage, the feasibility of the proposed method is analyzed through literature research and questionnaire survey; in the third stage, empirical research is carried out System development and testing are carried out through the method; the fourth step is to summarize and look forward to the results of the project.
The programming of clothing identification and management system is shown in Figure 2. In order to realize the clothing identification and management system, the front-end development tool of this subject is Android Studio, mainly using Java and XML languages, the back-end part is developed using the Python-based Django framework, and the technical connection between the front-end and back-end uses the C/S architecture (Client-Server) .
On the front end, an Android client usually consists of one or more basic components. When deciding which components to use to build an Android application, they need to be registered in the AndroidManifest.xml file, which is an XML configuration file in which application components and their characteristics and requirements can be declared, which can be summarized as XML The interface effect of APP is defined, and Java describes the behavioral logic of APP [12]. The web management terminal is developed by using the Python web framework, using HTML, CSS, and JS technology stacks.
On the back end, Django is a high-level open source model driven by the Python programming language. Using this architecture, you can easily and quickly create high-quality, easy-to-maintain, database-driven applications. This is also the main reason why many small and medium-sized entrepreneurial groups develop and adopt this architecture for design [12]. In addition, in the Django framework, you can also download many powerful third-party plug-ins and Python open source libraries, which make Django have strong scalability and gradually become the framework that web developers are keen to choose. In this topic, we will use The computer vision library OpenCV[13], which implements many common algorithms for image processing, provides the API interfaces required for clothing images, and uses Django to connect to the background database and provide various API interfaces to the Android front-end.
The technical connection of the front and back ends uses the C/S architecture (Client-Server), that is, the client and server structure. The client refers to the PC or terminal system used to send requests, and the server is used to provide services and information feedback for the client. The C/S structure can make full use of the advantages of the hardware environment at both ends, reasonably distribute the communication tasks to the client and server, and reduce the overhead of system communication implementation [14].
Figure 1 overall technology roadmap
Figure 2 Programming implementation diagram
V. Research work schedule
(1) From the 13th week of the seventh semester to the 19th week of the seventh semester, consult relevant literature on clothing identification to understand the research background, current situation and development trend of the topic; understand various methods of clothing identification; understand the implementation Identify specific methods of information management; plan the technical route of the subject, and formulate a reasonable work plan.
(2) From the 20th week of the seventh semester to the 01st week of the eighth semester, the front-end and back-end development tools will be configured, and the ink knife will set up the front-end pages and complete the construction of important front-end pages, which will be completed before the 01st week of the eighth semester Submit relevant work and materials related to the opening of the project.
(3) From the 2nd week of the eighth semester to the 08th week of the eighth semester, complete all the design of the front-end interface (apparel identification technology, identification information upload, identification information management), complete all the data storage of the MYSQL back-end database, and complete the front-end and back-end Data interaction enables the developed APP to run normally. Among them, the mid-term inspection task will be completed in the 05th week of the eighth semester.
(4) From the 09th week of the eighth semester to the tenth week of the eighth semester, improve and perfect the APP according to the test results and the opinions of the instructor, and write the first draft of the graduation thesis as required.
(5) From the 11th week of the eighth semester to the 11th week of the eighth semester, submit the graduation thesis and related materials for the completion of the project according to the requirements of the college for expert review.
(6) The twelfth week of the eighth semester - the twelfth week of the eighth semester, according to the review opinions of the college review expert group, revise the thesis;
prepare for the defense of the graduation project and make a defense PPT.
(7) From the 13th week of the eighth semester to the 13th week of the eighth semester, complete the system demonstration and graduation defense as required.
(8) From the 14th week of the eighth semester to the 16th week of the eighth semester, submit relevant thesis materials for completion to the instructor as required.
6. Main References
[1] China Electronic Commerce Research Center. 2015-2016 Annual Report on China's Garment E-commerce Industry [R]. 2016. [
2] Lu Jian, Liu Xin. Exports increased by 8.9% year-on-year. 2022 1-2 Inventory of China's Textile and Clothing Export Key Markets [J]. Textile and Clothing Weekly, 2022 (14): 12. [3]
Zhou Wenbo. Research on Garment Image Recognition, Positioning and Retrieval Based on Deep Learning [D]. Guangdong University of Technology, 2020
[4] Liu Hongwen, Li Cuiyun, Huang Zhigao, etc. The gap between consumer demand and product design supply of new Chinese clothing products [J]. Textile Journal, 2021,42(08):167-174. [5] Yang Dongning, Zeng Ting
, Zhu Yanjie. Principles and Applications of Image Recognition Technology [J]. Electronic Technology and Software Engineering, 2020(01): 102-103. [6]
Zhao Yue, Wang Laihua, Wang Weisheng, Qiao Lijuan, Ruan Quan. Fusion backpropagation without reference Evaluation of Blurred Image Quality[J]. Computer Application and Software, 2022,39(09):248-254+306.
[7] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[C]// Advances in neural information processing systems.2012:1097-1105.
[8] Kiapour MH, Han X, Lazebnik S, et al. Where to Buy It: Matching Street Clothing Photos in Online Shops[C]// IEEE International Conference on Computer Vision. IEEE, 2015.
[9]Lin K, Yang HF, Liu KH, et al. Rapid clothing retrieval via deep learning of binarycodes and hierarchical search[C]//Proceedings of the 5th ACM on International Conference on Multimedia Retrieval. ACM, 2015: 499-502.
[10] Ge Qingqing. Customized Clothing Model under the Experience Economy [J]. Chemical Fiber and Textile Technology, 2022, 51(06): 92-94.
[11] Xie Yangying, Cheng Zeli, Sui Shuqian. The Development Status of Advanced Custom Clothing in China and Countermeasure Analysis[J]. Textile Report, 2021,40(6):33-34.
[12] Bloch J. Effective Java[M]. Piscataway, NJ: IEEE Press, 2009:12.
[13] Gao Huanchao. Based on Design and Implementation of Django's Online Learning System [J]. Computer Programming Skills and Maintenance, 2021 (08): 35-36+92. [14]
Du Cong. Talking about OpenCV Computer Vision Library [J]. Science and Technology Information, 2016, 14(28):7-8.
Resource download address : https://download.csdn.net/download/sheziqiong/87904742
Resource download address : https://download.csdn.net/download/sheziqiong/87904742