Convolutional-neural-network learning method of multi-scale progressive accumulation

Abstract

The invention relates to a convolutional-neural-network learning method of multi-scale progressive accumulation. The method can be widely applied to the fields of machine vision and artificial intelligence such as target detection, target classification and target recognition. Firstly, average pooling operations are employed by the method to construct a multi-scale image pyramid on input images; and then images of different scales are progressively sent into a convolutional neural network, the convolutional neural network is enabled to carry out learning on the multiple images of the differentscales with progressive increasing of network depth and carry out progressive feature accumulation, and thus feature learning ability of the convolutional neural network is improved.

Description

technical field [0001] The invention relates to the fields of machine vision and artificial intelligence, in particular to a multi-scale progressively cumulative convolutional neural network learning method, which can be applied to target detection, target classification and target recognition systems. Background technique [0002] Convolutional neural network is currently the most popular deep learning algorithm. In recent years, a large number of target detection, target classification and target recognition algorithms based on convolutional neural networks have emerged. The accuracy of these algorithms largely depends on the feature learning ability of convolutional neural networks. [0003] At present, most of the research on convolutional neural networks focuses on improving the feature learning ability by deepening the network depth. Many extremely deep convolutional neural networks have emerged, such as GoogleNet, ResNet, DenseNet, etc. These extremely deep convoluti...

AI Technical Summary

Problems solved by technology

It can be seen that the algorithm in [2] fails to integrate the image downsampling operation into the learning of the convolutional neural network, and it actually needs to train multiple convolutional neural networks at the same time, and the amount of calculation is too large

Images