摘要
深度学习是机器学习的一个次领域,是现今研究的一个崭新的方向。它的研究目标是使机器具有人类进行学习和获取知识的能力,达到对数据的高层次抽象。自2012年以来,深度学习在计算机视觉以及自然语言处理两个领域取得巨大的成功。其中,卷积神经网络是深度学习最为常见的一种算法模型,现已成为图像领域的研究热点。它的参数共享机制避免传统网络中特征提取和数据重建过程中所带来的繁琐过程。图像分类作为计算机视觉的核心,无论是现在还是未来,都是研究的热点方向。使用CIFAR-10公用数据集和利用迁移学习的方法改进的ResNet50深度学习模型进行训练和测试,最终,在验证集得到不错的分类识别效果。
Deep learning is a subfield of machine learning and a new direction in today’s research. Its research goal is to enable machines to have the ability to learn and acquire knowledge, and to achieve a high level of abstraction of data. Since 2012, deep learning has achieved great success in both computer vision and natural language processing. Among them, convolutional neural network is the most common algorithm model for deep learning, and has become a research hotspot in the field of image. Its parameter sharing mechanism avoids the cumbersome process brought about by feature extraction and data reconstruction in traditional networks. As the core of computer vision, image classification is the hotspot of research both now and in the future. The CIFAR-10 public data set and the ResNet50 deep learning model improved by the migration learning method are used for training and testing. Finally, the classification set is well recognized.
作者
庞丝丝
黄呈铖
PANG Si-si;HUANG Cheng-cheng(College of computer and Information Engineering,Nanning Normal University,Nanning 530299)
出处
《现代计算机》
2019年第23期40-44,共5页
Modern Computer
基金
基于深度学习的图像目标识别方法(No.201810603187)
