Two-dimensional virtual fitting method based on neural network

Abstract

The invention discloses a two-dimensional virtual fitting method based on a neural network. The method comprises the steps of inputting an original figure graph and a target clothing graph; extractinga human body profile diagram, a human body articulation point diagram and a human body analysis diagram; generating a target human body analysis graph through a codec network; enabling the convolutional neural network to generate a deformed clothing graph; enabling the codec network to generate a rough result graph and a clothing mask graph; and recombining the rough result graph and the deformedgarment graph through the garment mask graph to generate a final effect graph. According to the method, a two-dimensional picture deep learning algorithm is applied, and compared with expensive three-dimensional hardware acquisition equipment and three-dimensional calculation with a large calculation amount, the method has the advantages that the cost is low, and the efficiency is high; a neuralnetwork method is used, a target human body analytic graph is generated through a coding and decoding structure, the neural network can be instruct to retain characteristics of all parts of a person in an original picture to the maximum extent, the convolutional neural network is used for conducting garment deformation on a target garment graph, and texture information of the target garment graphcan be reserved to the maximum extent.

Description

Technical field [0001] The invention relates to the fields of computer vision and image processing, in particular to a neural network-based two-dimensional virtual fitting method. Background technique [0002] With the development of Internet technology, online shopping is becoming more and more popular. Compared with shopping in physical stores, online shopping has the advantages of a wide variety of products and convenient shopping. Among all commodity categories, clothing commodities occupy an important proportion. But online clothing shopping also has some drawbacks. Compared with shopping in physical stores, where you can change clothes in real time and check the effect of clothing, online clothing shopping cannot provide renderings for consumers themselves, and consumers cannot intuitively obtain the impact of clothing on their own image in real time. Virtual fitting technology can solve this problem by providing consumers with clothing renderings. [0003] Virtual...

AI Technical Summary

Problems solved by technology

The disadvantage of this invention is that the 3D modeling technology is difficult, the hardware cost is high, the 3D scanning technology is required, the applicability is low, and the program is relatively complicated

The disadvantage of this invention is that the combination of neural network and thin-plate spline transformation can easily cause loss of clothing texture and dislocation of clothing shape

Images