Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Three-dimensional semantic segmentation method based on channel attention and multi-scale fusion

A multi-scale fusion and semantic segmentation technology, which is applied in neural learning methods, instruments, biological neural network models, etc., can solve problems such as information loss, inability to make full use of 3D data, and result impact, and achieve the effect of improving segmentation results

Pending Publication Date: 2022-07-12
XIANGTAN UNIV
View PDF0 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Due to the occlusion problem of objects in real scenes, after projecting objects onto a two-dimensional plane, part of the information will be lost, and the spatial structure information contained in the three-dimensional data cannot be fully utilized. The selection of the projection plane will also have a certain impact on the results of the algorithm.

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 semantic segmentation method based on channel attention and multi-scale fusion
  • Three-dimensional semantic segmentation method based on channel attention and multi-scale fusion
  • Three-dimensional semantic segmentation method based on channel attention and multi-scale fusion

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0085] The dataset used in this example is the S3DIS dataset, which is collected from the indoor environments of three different buildings and contains 271 rooms in 6 areas. It has a total of 695,878,620 point clouds, each of which has corresponding coordinates and color information, as well as semantic labels such as chairs, tables, floors, walls, etc., with a total of 13 categories. In this embodiment, regions 1, 2, 3, 4, and 6 are selected for training, and region 5 is selected for testing. During training, this embodiment samples the input points into a uniform number of 4096 points, and uses all the points during testing.

[0086] This example trains 150 epochs on two GeForce RTX 2080Ti GPUs with a batch size of 16, using an SGD optimizer with an initial learning rate of 0.05, momentum of 0.9, and weight decay rate of 10 -4 , and implemented on the Pytorch platform using Linux. After using the training set to train the network to obtain the model, the model performance ...

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 belongs to the technical field of three-dimensional point cloud data processing, and discloses a three-dimensional point cloud semantic segmentation method based on channel attention and multi-scale fusion. Firstly, to-be-segmented point cloud data is read, preprocessed and then input into a segmentation network; and then sequentially passing through four modules consisting of an encoder and a channel attention layer, wherein the encoder comprises a down-sampling layer, a grouping layer and a position self-adaptive convolution. Then, a multi-scale convolution context module is used for extracting point cloud context information, and finally, the point cloud context information sequentially passes through four decoders composed of upper sampling layers and unit Point Net networks. And the final segmentation result is obtained through a full connection layer with the size of k (category number). According to the method, the position information of the point cloud is fully utilized, the channel attention layer is introduced to re-calibrate the point cloud features in the channel dimension, more attention is paid to channel information useful for the segmentation task, a multi-scale convolution context module is further provided, and the segmentation task is more accurate. The features of different scales are captured in parallel by adopting cavity convolution with the same expansion rate and different kernel sizes, so that the segmentation result is improved.

Description

technical field [0001] The invention belongs to the technical field of three-dimensional point cloud data processing, and in particular relates to a three-dimensional point cloud semantic segmentation method based on channel attention and multi-scale fusion. Background technique [0002] With the development and rise of artificial intelligence technology, 3D point cloud data analysis has attracted widespread attention. Compared with two-dimensional images, 3D point clouds contain richer three-dimensional spatial information and are not affected by external factors such as illumination and perspective. More accurate and comprehensive characterization of the model. As the key content of scene understanding, 3D point cloud segmentation is one of the frontier research directions of artificial intelligence, and it is widely used in the fields of robotics, virtual reality, autonomous driving, and laser remote sensing measurement. [0003] Point cloud segmentation methods can be d...

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): G06V10/26G06V10/80G06V10/764G06V10/82G06K9/62G06N3/04G06N3/08
CPCG06N3/084G06N3/048G06N3/045G06F18/2415G06F18/253
Inventor 张莹孙月张露露王玉
Owner XIANGTAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products