摘要
图像识别和深度学习是目前人工智能领域应用较为广泛的算法,但在图像数据受损的情况下,图像识别的准确度和效率极大降低。文章从搭建深度学习框架入手,应用卷积神经网络并行架构,通过交替无监督和有监督学习训练网络,实现超分辨率重构目标图像,最后通过实验对图像修复效果进行验证。
Image recognition and deep learning are widely used algorithms in the field of artificial intelligence, but in the case of image data damage, the accuracy and efficiency of image recognition are greatly reduced. In this paper, starting from the construction of deep learning framework, the convolutional neural network parallel architecture is applied,and the super-resolution reconstruction of the target image is realized by alternating unsupervised and supervised learning training network, finally, the effect of inpainting is verified by experiments.
作者
范新刚
Fan Xingang(Guangzhou City Construction College,Guangzhou 510925,China)
出处
《江苏科技信息》
2020年第8期47-49,共3页
Jiangsu Science and Technology Information
关键词
图像识别
深度学习
卷积神经网络
image recognition
deep learning
convolutional neural network
