Multi-scale widened residual network, small target recognition and detection network and optimization method of small target recognition and detection network

Abstract

The invention belongs to the field of machine learning, and particularly relates to a multi-scale widened residual network, a small target recognition and detection network and an optimization methodthereof. The multi-scale widened residual network comprises a multi-scale widened convolutional layer and a multi-scale widened residual network unit structure which are connected in series; the multi-scale widened convolution layer comprises a plurality of convolution kernels of different scales which are arranged in parallel, and the output of the multi-scale widened convolution layer is the combination of the extracted features of the convolution kernels of different scales; the multi-scale widening residual network unit structure comprises a plurality of multi-scale widening convolution layers which are arranged in series, and skip layer connection between the two multi-scale widening convolution layers is set, so that output features of the two multi-scale widening convolution layersare directly subjected to maximum fusion. Small target recognition detection is carried out based on the obtained features, and the accuracy of small target object recognition can be improved.

Description

technical field [0001] The invention belongs to the field of machine learning, and in particular relates to a multi-scale widening residual network, a small target recognition and detection network and an optimization method thereof. Background technique [0002] With the development of computer vision, especially since the popularization and use of deep learning network models, object detection technology has made great progress. However, for the detection and image classification of low-pixel and small target objects, there is still a lack of effective detection and recognition algorithms. The main reason is that the structural design of the network and the information loss caused by the increase in the number of layers have a huge impact on small target detection. [0003] Convolutional neural networks have a variety of models, and their performance is gradually improving. There are basically three conventional strategies to improve network performance. One is to build n...

AI Technical Summary

Problems solved by technology

However, for the detection and image classification of low-pixel and small target objects, there is still a lack of effective detection and recognition algorithms. The main reason is that the structural design of the network and the information loss caused by the increase in the number of layers have a huge impact on small target detection.

Images