Please use this identifier to cite or link to this item:
Title: Image quality assessment based label smoothing in deep neural network learning
Authors: Chen, Zhou
Keywords: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
Issue Date: 2018
Abstract: For many computer vision problems, the deep neural networks are trained and validated based on the assumption that the input images are pristine (i.e., artifact-free). However, digital images are subject to a wide range of distortions in real application scenarios, while the practical issues regarding image quality in high level visual information understanding have been largely ignored. In this paper, in view of the fact that most widely deployed deep learning models are susceptible to various image distortions, the distorted images are involved for data augmentation in the deep neural network training process to learn a reliable model for practical applications. In particular, an image quality assessment based label smoothing method, which aims at regularizing the label distribution of training images, is further proposed to tune the objective functions in learning the neural network. Experimental results show that the proposed method is effective in dealing with both low and high quality images in the typical image classification task.
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:SCSE Research Reports (Staff & Graduate Students)

Files in This Item:
File Description SizeFormat 
  Restricted Access
281.15 kBAdobe PDFView/Open

Page view(s) 50

checked on Oct 26, 2020

Download(s) 50

checked on Oct 26, 2020

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.