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

Neural network automatic design system and method

An automatic design, neural network technology, applied in the field of visual odometer, can solve the problem that NAS cannot process time series information, and achieve the effect of good performance

Pending Publication Date: 2020-12-08
HANGZHOU WEIMING XINKE TECH CO LTD +1
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] What the present invention aims to solve is the technical problem that the existing NAS cannot process timing information

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
  • Neural network automatic design system and method
  • Neural network automatic design system and method
  • Neural network automatic design system and method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0035] A neural network automatic design system, including a hypernetwork structure and a controller model, such as figure 1 As shown, the super network structure includes a stem initial processing module, multiple block modules, and a tail module. The Stem module contains a convolution operation with a convolution kernel size of 7x7 and a stride (step size) of 2. The Tail module contains a normal convlstm layer for final processing of timing information. There are three types of block operations: the first one, convolution block, which contains a combination of multiple parallel convolution operations with different parameters; the size of the feature map output by the module and the number of channels are consistent with the input, and the number of channels (each layer volume The number of accumulated cores) remains unchanged. The second type, the reduction block, contains a combination of parallel convolution-based downsampling operations; the output feature map of this ...

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 neural network automatic design system and method, which utilize NAS (Network Attached Storage) to automatically design a model with a supervised VO (Virtual Office) task andimprove the NAS field and the VO field. For the NAS field, a more universal NAS framework is provided, and space and time sequence information can be processed at the same time so as to be suitable for video vision tasks. In the aspect of VO, a VONAS algorithm is used for searching to obtain a network model with better performance and lighter weight.

Description

technical field [0001] The invention belongs to the field of visual odometer in the field of computer vision, in particular to a neural network automatic design system and method. Background technique [0002] Visual Odometry (VO for short), as a key task in the field of autonomous driving and robotics, aims to estimate camera pose from consecutive frames. The traditional VO task is a typical geometric task, and the pose is obtained by strictly calculating the matching of feature points or pixels. With the rapid development of CNN (Convolutional Neural Networks, convolutional neural network) and RNN (recurrent neural network, recurrent neural network) in vision tasks, more and more end-to-end network frameworks are also applied to VO tasks. [0003] In the deep learning-based framework, the VO task is a video regression task, which is clearly different from semantic-based vision tasks (such as image classification, object detection). First, the VO task predicts a 6-DoF (de...

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): G06F30/27G06K9/00G06K9/62G06N3/04G06Q10/06
CPCG06F30/27G06Q10/06393G06V20/49G06N3/045G06F18/214
Inventor 蔡行张兰清李承远李宏
Owner HANGZHOU WEIMING XINKE TECH CO LTD