作者: Olivier Van Laere , Steven Schockaert , Bart Dhoedt
DOI: 10.1007/978-3-642-15951-0_32
关键词:
摘要: We explore the task of determining geographic location photos on Flickr, using combined evidence from Naive Bayes classifiers that are trained at different spatial resolutions. In particular, we estimate Flickr photos, based their tags, four scales, ranging a city-level granularity to fine-grained intracity areas. Using Dempster-Shafer's theory, combine output into single mass assignment. demonstrate experimentally induced belief and plausibility measures useful determine whether there is sufficient classify photo given granularity. Thus an adaptive method obtained, by which georeferenced most appropriate resolution.