Unlock instant, AI-driven research and patent intelligence for your innovation.

Three-dimensional model retrieval method adopting Siamese structure-based bidirectional long-short-term memory network

A long-short-term memory, three-dimensional model technology, applied in the field of image processing

Inactive Publication Date: 2020-04-10
HARBIN INST OF TECH
View PDF5 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The present invention proposes a multi-view 3D model retrieval method based on a Siamese-structured bidirectional long-short-term memory network to solve the fusion problem of multi-view features of a 3D model. Through the Siamese-structure long-short-term memory fusion view information via the network, the accuracy of the multi-view 3D model is enhanced. Retrieval accuracy

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Three-dimensional model retrieval method adopting Siamese structure-based bidirectional long-short-term memory network
  • Three-dimensional model retrieval method adopting Siamese structure-based bidirectional long-short-term memory network
  • Three-dimensional model retrieval method adopting Siamese structure-based bidirectional long-short-term memory network

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach 1

[0056] Specific implementation mode 1: The specific implementation mode of the present invention will be further elaborated in conjunction with the accompanying drawings. The multi-view three-dimensional model retrieval method based on the Siamese structure bidirectional long-short-term memory network proposed by the present invention utilizes the Siamese-structured cyclic neural network, adopts the independent perspective classification network to extract the single-view feature of the three-dimensional model, and introduces attention into the two-way long-short-term memory neural network. The force machine system uses a cyclic neural network to fuse single-view convolution features, and trains the network in the form of a large number of paired samples, which greatly improves the retrieval performance of multi-view 3D models.

[0057] Such as figure 1 Shown is the flow chart of the three-dimensional model retrieval method based on the Siamese structure bidirectional long-sho...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a three-dimensional model retrieval method adopting a Siamese structure-based bidirectional long-short-term memory network, belongs to the technical field of image processing,and aims to solve the problem of the fusion of the multi-view features of a three-dimensional model. According to the method of the invention, view information is fused through the Siamese structure-based bidirectional long-short-term memory network, so that the retrieval accuracy of the multi-view three-dimensional model can be enhanced. The method includes the following steps that: a convolutional neural network is constructed; a bidirectional long-short-term memory network is constructed; an attention mechanism is introduced to enhance the expression of model features; a Siamese structure-based bidirectional long-short-term memory network is constructed, multi-view information is fused; and the parameters of the Siamese structure-based bidirectional long-short-term memory network are trained, and the multi-view three-dimensional model is retrieved. With the method of the invention adopted, the retrieval of the three-dimensional model with multi-view expression can be realized. The method has a wide application prospect in the fields of three-dimensional model classification and retrieval.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a multi-view three-dimensional model retrieval method for multi-view feature fusion. Background technique [0002] At present, artificial intelligence is developing from perception to cognitive reasoning. 3D model retrieval plays an important role in promoting theoretical development and technological progress in the fields of 3D scene understanding, augmented reality, and robotics. In recent years, the new deep convolutional network extracts the bottom layer, middle layer, and high layer features in the image through abstraction, and combines the classifier to form an end-to-end network structure. This method has a strong visual expression ability and can efficiently extract model semantic information. However, natural images and data of 3D models represented by multiple views belong to different domains, and it is not suitable to directly handle multi-view 3D model retr...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06F16/53G06F16/55G06F16/583G06N3/04
CPCG06F16/53G06F16/583G06F16/55G06N3/045
Inventor 王滨王栋柳强赵京东刘宏
Owner HARBIN INST OF TECH