A Comprehensive Review of One-stage Networks for Object Detection
Yifan Zhang, Xu Li, Fei–Yue Wang, Baoguo Wei, Lixin Li
Abstract
Object detection has always been a hot topic in image processing, which is important in a variety of applications. With the advent of the era of big data and the continuous improvement of hardware computing power, deep learning gets more attention in object detection. One popular branch is regression-based (One-stage) model, which uses a single neural network to directly predict bounding boxes and class probabilities from the entire image by one evaluation. One-stage networks can effectively increase the detection speed. This article mainly describes object detection methods based on regression object detectors (One-stage methods), such as You Only Look Once (YOLO) series and Single Shot Multibox Detector (SSD) series. Then, their applications are briefly introduced. The development trend and future development direction of this type of object detection are discussed in the end.