摘要
目标检测作为计算机视觉领域的基础,其研究价值对于推动人工智能发展具有重要意义。长期以来,许多学者都致力于提升目标检测算法效率及性能的研究,无锚点(anchor-free)的目标检测深度学习算法以尺度灵活、鲁棒性强等优势,开始逐渐广泛应用于目标检测任务。介绍了目标检测领域中卷积神经网络和Transformer两种经典的网络架构;以核心网络架构为分类标准,分别介绍了基于卷积神经网络和基于Transformer的anchor-free目标检测深度学习算法,总结了这些算法的改进点和优缺点,并对该方向的未来发展及应用做出展望。
As the foundation of computer vision,object detection is of great significance to the development of artificial intelligence.For a long time,many scholars are committed to improve the efficiency and performance of object detection algorithms.Due to the advantages of flexible scale and strong robustness,deep learning algorithms of anchor-free object detection gradually are widely used in object detection.Convolutional neural network and Transformer,as the classical network architectures in the field of object detection,are introduced.Based on the core network architecture as the classification standard,the anchor-free object detection deep learning algorithms are introduced based on convolutional neural network and Transformer,respectively.The improvements,advantages and disadvantages of these algorithms are summarized,the future development and application of this direction are anticipated.
作者
陈恒星
刘一鸣
Chen Hengxing;Liu Yiming(Business School,Macao University of Science and Technology,Macao 999078,China;Business School,Sun Yat-sen University,Guangzhou 510006,China)
出处
《机电工程技术》
2024年第8期7-12,共6页
Mechanical & Electrical Engineering Technology
