Files in this item

FilesDescriptionFormat

application/pdf

application/pdfECE499-Sp2017-cheng-Bowen.pdf (1MB)Restricted to U of Illinois
(no description provided)PDF

Description

Title:Very Low-Quality Recognition Using Conventional Neural Network: With an Application to Face Identification
Author(s):Cheng, Bowen
Contributor(s):Huang, Thomas S.
Subject(s):Convolutional neural network
Computer vision
Classification
Recognition
Face recognition
Abstract:Visual recognition from very low-quality images is an extremely challenging task with great practical values, due to the ubiquitous existence of quality distortions during image acquisition, transmission, or storage. While deep networks have been extensively applied to low-quality image restoration and high-quality image recognition tasks respectively, less has been done on the important problem of recognition from very low-quality images. I propose a degradation-robust pre-training method to jointly tune reconstruction and classification with comprehensive analysis on improving deep learning models along this direction. This jointly tuning leverages the power of pre-training similar to that of transfer learning and generalizes conventional unsupervised pre-training and data augmentation methods. I did extensive experiments on a number of diverse real-world datasets to validate the effectiveness of the proposed method and applied this method on face identification tasks.
Issue Date:2017-05
Genre:Other
Type:Text
Language:English
URI:http://hdl.handle.net/2142/97846
Date Available in IDEALS:2017-08-17


This item appears in the following Collection(s)

Item Statistics