Rapid target detection method based on self-adaptive convolution

Abstract

The invention relates to a rapid target detection method based on adaptive convolution, and the method comprises the steps: obtaining image data with a label, and forming a training set; constructinga target detection network based on an adaptive convolution module; training the proposed target detection network on the formed training set until the target detection network converges to obtain a trained target detection network; and detecting the image by using the trained target detection network, removing repeated results, and outputting a final result. By dynamically adjusting the parameters of the convolution filter, the expression capability of the detection network for extracting the features is improved, the irrelevant interference features are inhibited, the feature response of theinterested target is enhanced, the extracted features are more suitable for the detection requirement of the current scene, and the detection performance of the extracted features on the small-scaletarget in the complex scene is improved.

Description

technical field [0001] The invention relates to a fast target detection method based on adaptive convolution, belonging to the technical fields of digital image processing, target detection and deep learning. Background technique [0002] Object detection is a basic computer vision perception task, which has a wide range of applications in areas such as autonomous driving, face recognition, and traffic video surveillance. Therefore, improving the accuracy and speed of object detection is an important research problem. [0003] In recent years, with the vigorous development of deep learning technology, computer vision has made great breakthroughs, and many advanced visual perception algorithms have been proposed. Among them, target detection, as a basic task, has attracted the attention of many researchers, so a series of efficient detection algorithms have also been proposed. Existing detection algorithms can be roughly divided into two categories: two-stage methods and si...

AI Technical Summary

Problems solved by technology

The article [4] uses a multi-layer feature prediction method to deal with the target detection problem at different scales. The shallow features are used to detect small-scale targets, and the deep features are used to detect larger-scale targets. The speed and accuracy of this method They have all made good progress and received a lot of research and applications. However, due to the limited receptive field of the shallow network, its features contain a large number of simple low-level features and interference information, which have a greater impact on the positioning of small-scale targets. lead to poor detection performance at small scales

The article [5] is further improved on the basis of [4]. A symmetrical hourglass network is built by using the deconvolution layer, and the high-level features are fused with the low-level features, and the accuracy is further improved. However, due to the network is more It is complex, the sampling of the prediction frame is more intensive, and the algorithm takes a lot of time, so it is difficult to be applied in actual use

Images