Manifold learning network and computer vision image set classification method based on the same

Abstract

The invention relates to a manifold learning network and a computer vision image set classification method based on the manifold learning network, the manifold learning network includes an input layer, a full connection layer, a correction layer, a Riemannian pooling layer, a second logarithmic mapping layer and an output layer which are mainly used to classify the collected video sequences. A simple learning network is built on Riemannian manifold, and two-way two-dimensional principal component analysis technology is used to optimize the weight parameters to complete the nonlinear feature extraction and classification of symmetric positive definite matrices. Compared with the prior art, the invention can effectively improve the object classification accuracy under the complex scene, andthe training time is short.

Description

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

AI Technical Summary

Benefits of technology

This patented technology allows us to learn about symmetrically positive set (SORM) matrices more easily than previous methods like Deep Learning or Neural Network techniques. It also includes various technical improvements such as simplifying the structure of networks by performing both bidirectional and 2D PCA processing instead of calculating critical values from specific directions, which makes them faster at computing compared to other models used previously. Additionally, this new algorithm does away with complicated mathematical operations that are commonly needed when working through SORM data sets due to its ability to efficiently process large amounts of data without losing important details during computation. Overall these benefits make our approach simpler and efficient overcomes challenges faced before implementing any current algorithms.

Problems solved by technology

This patented technical issue addressed in this patents relates to improving image classification tasks such as object detection (OCR) systems due to their flexibility in handling different kinds of images without requiring significant computational resources like convolution techniques. Specifically, there exist various ways to efficiently train models called multifocal networks which use only one picture per frame instead of all three. These approaches aim to reduce computation times while still achieving good performance across diverse datasets.

Images