A classification model training method, device, and classification method and device

Abstract

The application provides a classification model training method, device, and classification method and device. The classification model training method includes: obtaining a preset number of training image frames in a training video with labels; using a target neural network to perform feature learning on the training image frames, Extracting feature vectors for training image frames; based on the memory feature vectors corresponding to different classifications stored in the memory network, classify the training videos represented by the extracted feature vectors to obtain the classification results of the training videos; according to the classification results of the training videos and the training videos The comparison results between the labels of the target neural network are trained. The memory network has a strong memory ability, which can make up for the poor memory ability of the neural network itself. Regardless of the input sequence of any length, the memory network can memorize its characteristics well, so no matter what the input sequence is, it will not be affected. The recognition accuracy of the classification model will be affected due to the poor memory ability of the neural network.

Description

technical field [0001] The present application relates to the technical field of machine learning, in particular, to a classification model training method and device, and a classification method and device. Background technique [0002] Multi-layer feedback (Recurrent neural Network, RNN) neural network, also known as recurrent neural network, is an artificial neural network in which nodes are directional connected into a ring, which can use internal memory to process input sequences of any time sequence, and in video action prediction been widely applied. [0003] However, the memory ability of the RNN network is poor. When the input sequence is long, it cannot accurately classify the input sequence and cannot meet the use requirements for accuracy. Contents of the invention [0004] In view of this, the purpose of the embodiments of the present application is to provide a classification model training method and device, and a classification method and device, which can...

AI Technical Summary

Problems solved by technology

[0003] However, the memory ability of the RNN network is poor. When the input sequence is long, it cannot accurately classify the input sequence and cannot meet the use requirements for accuracy.

Images