Office Chair Components Recognition And Generation Based On Deep Learning

Office chair is used for daily work and social activities in the chair,which is mainly used by people in office work and desk supporting chair.At present,according to the similarities and differences of the shape on the market,the office chair can be divided into conference chair,middle class chair,staff chair,etc.,the office chair studied in this paper is mainly staff chair.In recent years,the office furniture industry has developed rapidly and the competition between them is also more intense,and the product appearance and function are more and more similar.This phenomenon makes the office chair industry from a blue sea into a red sea market.The appearance of the current office chair design has to have consistent with market choice,parts appearance remained unchanged for many years,and people become increasingly strong demand for personalized office chair,so need to seek a new way of design for the office chair form,which are helpful to the diversification of office chairs,and the production way of life of people have a positive and useful role.AI has long been a topic of concern in the science and technology circles,but in fact,ai technology has penetrated into every aspect of life,not just scientific research.At present,the research focus in the field of artificial intelligence has shifted to deep learning,which has greatly improved the development of many current artificial intelligence tasks and gradually become a possible new productivity tool in the design industry.Because deep learning is based on big data,it has the advantages of high efficiency and strong objectivity.It has played a great role in many fields(such as aircraft,intelligent robots,clothing,architecture,etc.)and successfully overcome many problems that modern computers cannot solve.In the field of furniture design,designers generally draw sketches by hand through experience,and computer is only his modeling tool.At present,artificial intelligence is seldom involved in the field of furniture design,especially in the field of office chair design.Office chair has a strong modular properties,and its module special clearly differentiated,by artificial intelligence technology to detection and segmentation existing office chair parts,thereby build office chair parts library,and based on the component library intelligent generating new office chair parts form the office chair parts material library,assist designer to design is valuable.The original intention of this study is to contribute to the practical application of intelligent design in the field of furniture,explore the intelligent identification and generation design method of office chair components based on deep learning,and improve the efficiency of design and scheme.In order to achieve the purpose of this study,the following research work is completed: Firstly,an image classification model is constructed based on Rse Net50 network,and the source data of web crawler is screened by the image classification model,and the image data of office chair and non-office chair are discriminated intelligently,so as to obtain the optimal data set of office chair image.Secondly,the basic components of the office chair are divided into modules,and the detection and segmentation model of office chair components is trained to establish the office chair component library.Mainly based on the product shape analysis method to decompose step by step for the office chair will chair the whole basic tectonic units(including a number of different categories of related parts),combining the expert group of card sorting office chair components for integrated deduction,will chair the basic tectonic units are divided into five modules: back of the chair,the back frame,seat cushion,armrest,chair leg.Using Py Torch deep learning framework,we compared the recognition accuracy,recognition reliability and inference speed of three common detection networks(YOLOv5 network,Faster RCNN network and SSD network)and segmentation networks(CGNet network,Unet network and Deep Labv3+ network).Finally,YOLOv5 network and Deep Labv3+network are selected to build the detection and segmentation model of office chair parts,so as to establish the office chair part library.Finally,intelligent assistant office chair modeling design is based on generative adversarial network.The process and characteristics of data-driven intelligent assistant design method for office chair modeling were proposed.DCGAN was trained to generate office chair component model,and an interface program was made based on Py Qt5 combined with the trained generation network model to facilitate the design.The typical scene of office chair design is analyzed and constructed.Taking designers’ participation in the overall design of office chair(seeking inspiration for armrest parts)as an example,the armrest picture is selected as the design inspiration to deepen the design,and the advantages and disadvantages of intelligent design are pointed out according to the generated results.The final result of this study is to propose a deep learning-based intelligent identification and generation design method for office chair components,and carry out design practice under the guidance of this method.This study also made some innovations on the basis of previous studies.The main innovations are as follows:First,the current cutting-edge technology is applied to the field of office chair design to assist designers in design.Second,the object detection and segmentation algorithm is applied to the construction of office chair component library.Third,DCGAN is used to generate new parts to form a material library of office chair parts,providing inspiration to designers.On the whole,the data-driven intelligent assisted office chair modeling design method constructed in this paper based on deep learning technology can provide inspiration source and certain design help for designers more efficiently,realize computer collaborative design,and provide a new idea for intelligent assisted furniture design.Its application prospect is very broad.