This work proposes a two-stage method that reconstructs the map of a scene from tagged
photographs of that scene. In the first stage, several methods are proposed that transform tag data from the photographs into an intermediary distance matrix. These methods are compared against each other. In the second stage, an approach based on the physical mass-spring system is proposed that transforms the distance matrix into a map. This approach is compared against and outperforms MDS-MAP(P) when given human tagged input photographs. Experiments are carried out on two test datasets, one with 67 tags, and the other with 19. An evaluation method is described and the optimal overall
reconstruction generates maps with accuracies of 47% and 66% respectively for the two
test datasets, both scoring roughly 40% higher than a random reconstruction. The map
reconstruction method is applied to three sample datasets and the resulting maps are
qualitatively evaluated.
Identifer | oai:union.ndltd.org:TORONTO/oai:tspace.library.utoronto.ca:1807/17150 |
Date | 24 February 2009 |
Creators | Appel, Ron |
Contributors | Aarabi, Parham |
Source Sets | University of Toronto |
Language | en_ca |
Detected Language | English |
Type | Thesis |
Format | 10962992 bytes, application/pdf |
Page generated in 0.0016 seconds