摘要
卷积神经网络在图像识别处理方面有着优秀的表现,但是只能处理单个输入,无法在多个输入之间建立联系。循环神经网络则在处理前后相关的序列信息上有着独特的优势。将两种神经网络算法联系起来,可以用于实现图像的语言序列描述,具体方法为:首先用卷积神经网络将图片的特征提取,后连接到LSTM模型,与输入的语言序列共同训练网络达到描述图像的目的。输入的数据应当根据需要做适当的预处理,以获得更好的表现。
Convolutional neural network has a satisfactory performance in image recognition and processing, but can only deal with a single input, cannot be established between multiple inputs. Recurrent neural network has a unique advantage in dealing with sequence information. The two kinds neural network algorithm can be used to achieve the image sequence of language description. Firstly, the convolution neural network is used to extract the feature of the image, and then connected to the LSTM model to train the network with the input language sequence to achieve the purpose of image caption. Input data should be properly pretreated as needed to achieve better performance.
出处
《电脑知识与技术》
2017年第8X期178-179,182,共3页
Computer Knowledge and Technology
关键词
循环神经网络
卷积神经网络
长短时间记忆模型
图像描述
数据预处理
Recurrent Neural Network
Convolutional Neural Network
Long Short-Term Memory
image caption
data preprocessing
