Research Advanced in the Object Detection Based on Deep Learning

Abstract

Object detection is one of the most basic tasks in the field of Computer vision, which targets to localize and allocate a wide range of predefined substances from images to their corresponding classification. Thanks to the rapid progress of deep learning, object detection algorithms based on convolutional neural networks have been applied in different fields, and have achieved breakthroughs both in accuracy and efficiency compared to traditional detection schemes. In this paper, based on detailed literature research and analysis, a comprehensive evaluation of object detection research advances are provided and specifically, we divide existing representative algorithms into three main frameworks, including traditional detection algorithms, anchor-based and anchor-free detection algorithms. We then conduct a series of experiments to analyze the achievement of different detection algorithms on some common datasets. Finally, we summarize the main challenges and provide an outlook on the future research directions of object detection.