Dartmouth logo Dartmouth College Computer Science
Technical Report series
CS home
TR home
TR search TR listserv
By author: A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
By number: 2017, 2016, 2015, 2014, 2013, 2012, 2011, 2010, 2009, 2008, 2007, 2006, 2005, 2004, 2003, 2002, 2001, 2000, 1999, 1998, 1997, 1996, 1995, 1994, 1993, 1992, 1991, 1990, 1989, 1988, 1987, 1986

Scene classification from degraded images: comparing human and computer vision performance
Tim M. Tadros
Dartmouth TR2017-820

Abstract: People can recognize the context of a scene with just a brief glance. Visual information such as color, objects and their properties, and texture are all important in correctly determining the type of scene (e.g. indoors versus outdoors). Although these properties are all useful, it is unclear which features of an image play a more important role in the task of scene recognition. To this aim, we compare and contrast a state-of-the-art neural network and GIST model with human performance on the task of classifying images as indoors or outdoors. We analyze the impact of image manipulations, such as blurring and scrambling, on computational models of scene recognition and human perception. We then create and analyze a measure of local-global information to represent how each perceptual system relies on local and global image features. Finally, we train a variety of neural networks on degraded images to attempt to build a neural network that emulates human performance on both classificaton accuracies and this local-global measure.

Note: Senior Honors Thesis. Advisor: Emily Cooper.


PDF PDF (5128KB)

Bibliographic citation for this report: [plain text] [BIB] [BibTeX] [Refer]

Or copy and paste:
   Tim M. Tadros, "Scene classification from degraded images: comparing human and computer vision performance." Dartmouth Computer Science Technical Report TR2017-820, May 2017.


Notify me about new tech reports.

Search the technical reports.

To receive paper copy of a report, by mail, send your address and the TR number to reports AT cs.dartmouth.edu


Copyright notice: The documents contained in this server are included by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a non-commercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.

Technical reports collection maintained by David Kotz.