Deep Learning-Based Lightweight Target Detection

Abstract

With the rapid development of computer vision technology, target detection techniques have also been widely used in various fields. However, traditional target detection methods usually require a large amount of computational resources and complex model structures, which pose challenges to the realization and deployment of target detection. Therefore, lightweight target detection methods arise. Lightweight target detection methods refer to reducing the computational and parametric quantities of the model as much as possible while keeping the detection accuracy unchanged. The emergence of this method can not only improve the speed and efficiency of target detection, but also make the target detection technology more adaptable to embedded devices and other resource-limited scenarios. In this paper, through the summary of the deep learning method, the impact on the accuracy and the reduction of computation is compared to give a reference to the subsequent embedded systems or small computational amount of system lightweighting.