A Spatial Information Learning Method Based on Artificial Neural Network

Abstract

The invention relates to a spatial information learning method based on an artificial neural network, belonging to the technical field of deep learning. Including: using the traditional neural network to abstract the input image, extracting the feature map in the middle and high layers of the traditional neural network to obtain the feature vector F; performing filtering and mapping operations on F to obtain LF to enhance the significance of the data; Convolution operation to obtain a more abstract feature expression X; perform GAP dimensionality reduction processing on X to obtain feature points X*; use the formula D=X for X* * ×X *T Perform correlation analysis to obtain a correlation matrix; perform projection operation on D to obtain the structural feature vector CD through the formula CD=D*V; obtain the final feature and output it through the formula Y(F)=F+λ*CD on CD. Compared with the prior art, the present invention can recognize the geometrically deformed object without relying on the data enhancement method, making the neural network more robust to the recognition of invariance.

Description

technical field [0001] The invention relates to a spatial information learning method based on an artificial neural network, which is a method for extracting invariant features by using a neural network, in particular to a method for extracting spatial information by using a deep convolutional neural network, and belongs to the technical field of deep learning. Background technique [0002] Convolutional Neural Network (CNN), as one of the best feature extractors at the present stage, not only has its expressiveness shine in the field of computer vision, but also the local perception principle of convolution kernel has made NLP, Go game (AlphaGo) There has been progress in the non-visual field. In order to extract more expressive features, more researchers are working on how to design an efficient convolutional neural network architecture. For example, Alexnet—it is the first leap-like result improvement of convolutional neural network in recent times, VGG, Inception—can ma...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to solve the problem that the artificial neural network relies on data preprocessing methods to improve the generalization ability of the model at the present stage, and proposes a spatial information learning framework, so that the features captured by the artificial neural network are invariant and reduce the number of artificial neural networks. Dependency on data preprocessing

Images