Three-dimensional model classification method based on end-to-end deep integrated learning network

Abstract

The present invention discloses a three-dimensional model classification method based on an end-to-end deep integrated learning network. The method comprises: using an end-to-end deep learning integration strategy to input a three-dimensional mesh model; extracting multi-view representation; establishing an integrated deep learning network containing a base-based learner and an integrated learner;and automatically extracting composite features of the three-dimensional model to complete model classification. Experiments show that the method disclosed by the present invention achieves classification accuracy of 96.04%, 92.79%, 98.33%, 98.44%, and 98.63% on the ModelNet10, ModelNet40, SHREC10, SHREC11, and SHREC15 data sets respectively, the result is significantly better than other multi-view classification algorithms, and the effectiveness of the method is also preliminarily verified.

Description

technical field [0001] The invention relates to the technical fields of computer graphics, computer vision and intelligent recognition, in particular to a three-dimensional model classification method based on an end-to-end deep integrated learning network (EnsembleNet). Background technique [0002] At present, with the continuous development of 3D modeling, scanning, and computer vision, the research and application of related technologies such as unmanned driving, 3D scene roaming, and smart city construction have attracted widespread attention. Among them, the effective identification of 3D models is the basic research problem. [0003] The construction of features and the selection of classification models are the key to determine the quality of classification. Especially for complex data types such as 3D models, the establishment of appropriate features is a hot topic for researchers in related fields, and it is also a research difficulty in the industry. Deep learni...

AI Technical Summary

Problems solved by technology

Since the outstanding advantage of deep learning is that it can complete feature self-learning; and this type of method has already carried out a primary feature extraction when inputting vector data, it is inevitable to lose the original information of the 3D model, and cannot make full use of the advantages of deep learning feature self-learning , affecting the effect of deep learning

Images