Intelligent Generation Of Basic Ceramic Shapes Based On A Single Plan Sketch

With the rise of VR/AR technology,the demand for 3D models is also increasing rapidly.At present,most of the 3D models are built manually by professional modelers through relevant modeling software,and this way of building models has problems such as cumbersome,time-consuming,and inefficient modeling processes.Moreover,nonprofessionals can only use a few abstract paintings to express their ideas,and cannot use concrete 3D models to express their ideas accurately.In view of this,this project investigates an algorithm to reconstruct a concrete 3D model using a single hand-drawn sketch of a ceramic vessel type.This paper designs a 3D reconstruction system for basic ceramic shapes to solve the problems of traditional manual 3D modeling.Firstly,an SNP2S-based ceramic sketch generation method is proposed to realize the transformation from a photograph to a handdrawn sketch of a ceramic vessel.It is demonstrated that the method is able to generate ceramic hand sketches with line details and line continuity,and the obtained FID(Fréchet Inception Distance)score is better than the AODA network structure.Then,we use the differentiable renderer Soft Ras to reconstruct the generated ceramic hand-drawn sketches in 3D and improve the single-view 3D reconstruction algorithm based on deep learning,which not only generates a 3D model that approximates the ceramic features but also improves the accuracy of the generated model.The experimentally obtained Io U(Intersection over Union)score and running time are also better than the Sketch2 model network structure.In addition,this paper proposes an optimization algorithm based on the Laplacian smoothing algorithm,Taubin smoothing algorithm,and loop mesh subdivision algorithm for 3D ceramic ware-shaped meshes,which optimizes the problem of generating ceramic meshes with a too-rough surface for the characteristics of ceramic ware-shaped.This paper researches and implements the single-view 3D reconstruction of ceramic vessel models based on Sketch2 Ceramic,which is an attempt to apply deep learning to the3 D reconstruction of ceramic vessel models.The application not only allows laymen to use abstract sketches to get concrete 3D ceramic models but also reduces the 3D mesh noise generated by the scribbled sketches,which can effectively solve the current problem of not being able to generate concrete 3D models from abstract ceramic hand-drawn sketches,largely reducing the tedious operations in the industrial modeling process and improving the modeling efficiency.