摘要
近几年,随着人工智能深度学习的不断发展,计算机视觉领域也逐渐发展扩大,先后出现了图像检索、图像自动标注等新的研究方向。最初为支持图像检索而逐渐兴起的图像自动标注技术,可以在一定程度上跨越"语义鸿沟",让计算机自动给图像加上能够反映图像内容的语义描述,从而减少人工标注成本。深度学习作为人工智能领域的新技术,其复杂的神经网络结构能够在学习到图像特征后快速输出结果,如果将深度学习应用于图像自动标注,将大大节约人工标注时间,降低人工标注成本。文章为探究深度学习在图像自动标注上的可行性,将以作者的生活照为样本数据,使用深度卷积神经网络与深度循环神经网络进行图像处理,最后输出图像的文字描述。
In recent years,with the continuous development of artificial intelligence deep learning,the field of computer vision has gradually developed and expanded,and new research directions such as image retrieval and automatic image annotation have emerged.The automatic image annotation technology,which was originally developed to support image retrieval,can cross the“semantic gap”to a certain extent,allowing the computer to automatically add a textual description of the image content to the image,thereby reducing the cost of manual labeling.As a new technology in the field of artificial intelligence,the complex neural network structure of deep learning can quickly output results after learning image features.If applied to automatic image annotation,deep learning will greatly save manual labeling time and reduce manual labeling cost.In order to explore the feasibility of deep learning in automatic image annotation,the article will take the author's photos of life as sample data,use deep convolutional neural network and deep recurrent neural network for image processing,and output the text description of the image.
作者
魏珺洁
WEI Junjie(College of Computer,Xi'an Shiyou University,Xi'an 710065,China)
出处
《智能计算机与应用》
2020年第3期111-113,118,共4页
Intelligent Computer and Applications
关键词
深度学习
深度卷积神经网络
深度循环神经网络
图像自动标注
deep learning
deep convolutional neural network
deep recurrent neural network
automatic image annotation
