SlideShare a Scribd company logo
1 of 29
Automatically generating keywords for georeferenced images Ross Purves1, Alistair Edwardes1, Xin Fan2, Mark Hall2 and Martin Tomko1 1University of Zurich 2University of Sheffield 3Cardiff University
What I want to talk about… How are images indexed? How do we describe images? How can we (semi?)-automaticallygenerate (geographically related) keywords that describe image content?
Image indexing Image indexing is crucial, since image search is almost always based on keywords Keywords basically assigned in four ways: Manual annotation Terms freely chosen by indexers (these might include tags nowadays) Terms selected from a controlled vocabulary Automatic annotation Terms extracted from text thought to be related to an image Terms assigned to an image on the basis of content-based techniques
…but… Image search (based on such annotation) often does not meet user expectations
The semantic gap… “…the lack of coincidence between the information that one can extract from the visual data and the interpretation that the same data have for a user in a given situation.”  (Smeulders et al., 2000)
…in other words Search fails because annotations are not the same as search terms… …unlike text search – we are using a proxy for content The challenge is, therefore, to develop methods that better match user expectations -> and are therefore more universal Taken up by the ESP game/ Google’s image labeller
How should we describe these pictures? Image Javier Corripio  Image Google Streetview
Theory of image description Panofsky-Shatford facet matrix – Shatford (1986)
The where facet In the Tripod project, we are especially interested in describing images based on where they were taken We suppose, that in the medium-term all images will be georeferenced Panofsky-Shatfordmatrix suggests some ways we might describe images I’m going to concentrate on where/generic of (Martin will talk tomorrow about one element of the where/ specific of)
Where/ generic of The generic of represents a kind of place Kinds may relate to basic levels Basic levels (e.g. Rosch, 1977) are terms used in natural language which are informative and summative (e.g. table vs. furniture or square table)  Basic levels are probably very good indexing terms For example: Mountains, valleys, desert, ravines Street, pavement, house Empirical research has explored what these basic levels are in human subject experiments and we explored them in UGC/VGI How can we generate terms which relate to these kinds automatically?
Basic process Start with an image associated with a coordinates and (sometimes) direction Basic process Identify potential visible area Query spatial data within visible area for candidate keywords Rank and filter candidate keywords to generate final keyword list
Identifying visible area Camera parameters extracted from EXIF Content-based check for building combined with landcover data to determine urban or rural case – controls range of viewshed If no direction information, 360° viewshedgenerated, otherwise sector defined on basis of camera parameters
Data: OpenStreetMap, ©SwissTopo and ©Ordnance Survey
Basic keywording process Identify available data in region Query data using viewshed for data classes Map data classes to (potentially multiple) concepts – remove duplicates Concepts expanded to multiple (potentially multi-lingual) candidate keywords Rank candidate keywords according to area, probabilistic salience and web salience
Initial result ,[object Object]
Spotted feature types:
Corine:0_wcsNodata, 312_coniferousForest, 322_moorsAndHeathland, 512_waterBodies
Strategi:5610, 5250, 5385
VectorMap:PrivateRoadRestrictedAli, MinorRoadAlignment, UnimprovedGrass, HeathlandAndUnimprovedGrass, ConiferousWoodland, WaterBodies, RiverWaterPolygon, Lakesselect, MixedWoodland, BroadleafedWoodland, WaterFeature, BuildingPolygon,[object Object]
Initial filtering  ,[object Object]
Duplicates removed
Returned ranked by area in viewshedWoodland, Lakes, NaturalGrassland, Reservoir, SingleBuiltWorksByFunction, Tracks, Moor, Streets, Stream, Pond, Forest, ConiferousTrees, Tarn, Heath, Byways
Final filtering and ranking Ranking and filtering based on spatial extent, descriptive and web salience Spatial extent: concepts covering large area typically important (but favours landcover related concepts) Descriptive salience: weight rare concepts with respect to surroundings higher than common (e.g. a village shop is more salient than one on a high street) Web salience: Query web with keywordsand local toponymsto find common combinations
Keywords:Reservoir, Forest, Moors, Woodland
Keywords: Lake, Meadows,  Settlement, Reservoir, Forest
Keywords:Street, Footpath, Gallery, Steps, Parkland, Broadway
Are we any good overall? Initial results (with an interim solution) – about 40% of keywords are good or very good for a set of 20 images, assessed by ~70 users
Where are we? We can (semi) automatically annotate images based on their location, with keywords related to geography Concept ontology allows us to switch data providers in and out easily Filtering and ranking reduces initial long list of candidate concepts to set of keywords ~40% of keywords are good/very good
Some lessons learned Many devices still relatively error prone (particularly with respect to direction) Open Street Map has very rich (but sometimes esoteric) attribution – but richness in urban areas very beneficial National Mapping Agency data essential in rural areas General classes (e.g. Building) difficult to use (either too many or too general keywords)

More Related Content

Similar to 4B_3_Automatically generating keywods for georeferenced imaged

Experimental categorization and deep visualization
 Experimental categorization and deep visualization Experimental categorization and deep visualization
Experimental categorization and deep visualizationEverardo Reyes-García
 
Searching in high dimensional spaces index structures for improving the perfo...
Searching in high dimensional spaces index structures for improving the perfo...Searching in high dimensional spaces index structures for improving the perfo...
Searching in high dimensional spaces index structures for improving the perfo...unyil96
 
UCT GIS Labs
UCT GIS LabsUCT GIS Labs
UCT GIS Labspvhead123
 
Semantic Integration at large scale
Semantic Integration at large scaleSemantic Integration at large scale
Semantic Integration at large scaleGrowSmarter
 
Towards preservation of semantically enriched architectural knowledge
Towards preservation of semantically enriched architectural knowledgeTowards preservation of semantically enriched architectural knowledge
Towards preservation of semantically enriched architectural knowledgeStefan Dietze
 
Argumentation Trails and Topic Maps
Argumentation Trails and Topic MapsArgumentation Trails and Topic Maps
Argumentation Trails and Topic MapsLutz Maicher
 
University of Nottingham - NGI Geospatial Science Example Activities
University of Nottingham - NGI Geospatial Science Example ActivitiesUniversity of Nottingham - NGI Geospatial Science Example Activities
University of Nottingham - NGI Geospatial Science Example ActivitiesJeremy Morley
 
2-3-intro11.ppt
2-3-intro11.ppt2-3-intro11.ppt
2-3-intro11.pptTemur10
 
Dig the new breed: how open approaches can empower archaeologists
Dig the new breed: how open approaches can empower archaeologistsDig the new breed: how open approaches can empower archaeologists
Dig the new breed: how open approaches can empower archaeologistsDART Project
 
Quality, Relevance and Importance in Information Retrieval with Fuzzy Semanti...
Quality, Relevance and Importance in Information Retrieval with Fuzzy Semanti...Quality, Relevance and Importance in Information Retrieval with Fuzzy Semanti...
Quality, Relevance and Importance in Information Retrieval with Fuzzy Semanti...tmra
 
Ben Shneiderman: Thrill of Discovery
Ben Shneiderman: Thrill of DiscoveryBen Shneiderman: Thrill of Discovery
Ben Shneiderman: Thrill of Discoveryruss9595
 
Semantics for visual resources: use cases from e-culture
Semantics for visual resources: use cases from e-cultureSemantics for visual resources: use cases from e-culture
Semantics for visual resources: use cases from e-cultureGuus Schreiber
 
Citizen Science Involving Collections of Standardized Community Data
Citizen Science Involving Collections of Standardized Community DataCitizen Science Involving Collections of Standardized Community Data
Citizen Science Involving Collections of Standardized Community DataWerner Leyh
 
How Flickr Helps us Make Sense of the World
How Flickr Helps us Make Sense of the WorldHow Flickr Helps us Make Sense of the World
How Flickr Helps us Make Sense of the Worldmor
 
Big Data Palooza Talk: Aspects of Semantic Processing
Big Data Palooza Talk: Aspects of Semantic ProcessingBig Data Palooza Talk: Aspects of Semantic Processing
Big Data Palooza Talk: Aspects of Semantic ProcessingNa'im Tyson
 
Human Terrain Analysis at George Mason University (DAY 1)
Human Terrain Analysis at George Mason University (DAY 1)Human Terrain Analysis at George Mason University (DAY 1)
Human Terrain Analysis at George Mason University (DAY 1)Rich Heimann
 

Similar to 4B_3_Automatically generating keywods for georeferenced imaged (20)

Experimental categorization and deep visualization
 Experimental categorization and deep visualization Experimental categorization and deep visualization
Experimental categorization and deep visualization
 
Searching in high dimensional spaces index structures for improving the perfo...
Searching in high dimensional spaces index structures for improving the perfo...Searching in high dimensional spaces index structures for improving the perfo...
Searching in high dimensional spaces index structures for improving the perfo...
 
UCT GIS Labs
UCT GIS LabsUCT GIS Labs
UCT GIS Labs
 
Learn about Your Location (Using ALL Your Data)
Learn about Your Location (Using ALL Your Data)Learn about Your Location (Using ALL Your Data)
Learn about Your Location (Using ALL Your Data)
 
Semantic Integration at large scale
Semantic Integration at large scaleSemantic Integration at large scale
Semantic Integration at large scale
 
Towards preservation of semantically enriched architectural knowledge
Towards preservation of semantically enriched architectural knowledgeTowards preservation of semantically enriched architectural knowledge
Towards preservation of semantically enriched architectural knowledge
 
Argumentation Trails and Topic Maps
Argumentation Trails and Topic MapsArgumentation Trails and Topic Maps
Argumentation Trails and Topic Maps
 
University of Nottingham - NGI Geospatial Science Example Activities
University of Nottingham - NGI Geospatial Science Example ActivitiesUniversity of Nottingham - NGI Geospatial Science Example Activities
University of Nottingham - NGI Geospatial Science Example Activities
 
2-3-intro11.ppt
2-3-intro11.ppt2-3-intro11.ppt
2-3-intro11.ppt
 
40120140501013
4012014050101340120140501013
40120140501013
 
Dig the new breed: how open approaches can empower archaeologists
Dig the new breed: how open approaches can empower archaeologistsDig the new breed: how open approaches can empower archaeologists
Dig the new breed: how open approaches can empower archaeologists
 
13584 27 multimedia mining
13584 27 multimedia mining13584 27 multimedia mining
13584 27 multimedia mining
 
Visual Search
Visual SearchVisual Search
Visual Search
 
Quality, Relevance and Importance in Information Retrieval with Fuzzy Semanti...
Quality, Relevance and Importance in Information Retrieval with Fuzzy Semanti...Quality, Relevance and Importance in Information Retrieval with Fuzzy Semanti...
Quality, Relevance and Importance in Information Retrieval with Fuzzy Semanti...
 
Ben Shneiderman: Thrill of Discovery
Ben Shneiderman: Thrill of DiscoveryBen Shneiderman: Thrill of Discovery
Ben Shneiderman: Thrill of Discovery
 
Semantics for visual resources: use cases from e-culture
Semantics for visual resources: use cases from e-cultureSemantics for visual resources: use cases from e-culture
Semantics for visual resources: use cases from e-culture
 
Citizen Science Involving Collections of Standardized Community Data
Citizen Science Involving Collections of Standardized Community DataCitizen Science Involving Collections of Standardized Community Data
Citizen Science Involving Collections of Standardized Community Data
 
How Flickr Helps us Make Sense of the World
How Flickr Helps us Make Sense of the WorldHow Flickr Helps us Make Sense of the World
How Flickr Helps us Make Sense of the World
 
Big Data Palooza Talk: Aspects of Semantic Processing
Big Data Palooza Talk: Aspects of Semantic ProcessingBig Data Palooza Talk: Aspects of Semantic Processing
Big Data Palooza Talk: Aspects of Semantic Processing
 
Human Terrain Analysis at George Mason University (DAY 1)
Human Terrain Analysis at George Mason University (DAY 1)Human Terrain Analysis at George Mason University (DAY 1)
Human Terrain Analysis at George Mason University (DAY 1)
 

More from GISRUK conference

7B_3_Matterhorn on the horizon
7B_3_Matterhorn on the horizon7B_3_Matterhorn on the horizon
7B_3_Matterhorn on the horizonGISRUK conference
 
7B_2_Topological consistent generalization of openstreetmap
7B_2_Topological consistent generalization of openstreetmap7B_2_Topological consistent generalization of openstreetmap
7B_2_Topological consistent generalization of openstreetmapGISRUK conference
 
7A_4_Gps data collection setting for pedestrian activity modelling
7A_4_Gps data collection setting for pedestrian activity modelling7A_4_Gps data collection setting for pedestrian activity modelling
7A_4_Gps data collection setting for pedestrian activity modellingGISRUK conference
 
5A_3_GIS based spatial modelling for improving the sustainability of aggregat...
5A_3_GIS based spatial modelling for improving the sustainability of aggregat...5A_3_GIS based spatial modelling for improving the sustainability of aggregat...
5A_3_GIS based spatial modelling for improving the sustainability of aggregat...GISRUK conference
 
5A_1_Land evaluation techniques comparing fuzzy ahp with ideal point methods
5A_1_Land evaluation techniques comparing fuzzy ahp with ideal point methods5A_1_Land evaluation techniques comparing fuzzy ahp with ideal point methods
5A_1_Land evaluation techniques comparing fuzzy ahp with ideal point methodsGISRUK conference
 
4B_1_How many volunteers does it take to map an area well
4B_1_How many volunteers does it take to map an area well4B_1_How many volunteers does it take to map an area well
4B_1_How many volunteers does it take to map an area wellGISRUK conference
 
4A_1_Uncertainty in the 2001 output area classification for the census of eng...
4A_1_Uncertainty in the 2001 output area classification for the census of eng...4A_1_Uncertainty in the 2001 output area classification for the census of eng...
4A_1_Uncertainty in the 2001 output area classification for the census of eng...GISRUK conference
 
3A_4_Applying network analysis to quantify accessibility to urban greenspace ...
3A_4_Applying network analysis to quantify accessibility to urban greenspace ...3A_4_Applying network analysis to quantify accessibility to urban greenspace ...
3A_4_Applying network analysis to quantify accessibility to urban greenspace ...GISRUK conference
 
3A_2_Modelling health-harming behaviours in a socially ranked geographic space
3A_2_Modelling health-harming behaviours in a socially ranked geographic space3A_2_Modelling health-harming behaviours in a socially ranked geographic space
3A_2_Modelling health-harming behaviours in a socially ranked geographic spaceGISRUK conference
 
1A_3_A geodemographic classification of london primary schools
1A_3_A geodemographic classification of london primary schools1A_3_A geodemographic classification of london primary schools
1A_3_A geodemographic classification of london primary schoolsGISRUK conference
 
UK Map Challenge Aidan Slingsby
UK Map Challenge   Aidan SlingsbyUK Map Challenge   Aidan Slingsby
UK Map Challenge Aidan SlingsbyGISRUK conference
 
SP_4 Supporting spatial negotiations in land use planning
SP_4 Supporting spatial negotiations in land use planningSP_4 Supporting spatial negotiations in land use planning
SP_4 Supporting spatial negotiations in land use planningGISRUK conference
 
SP_3 Automatic identification of high streets and classification of urban lan...
SP_3 Automatic identification of high streets and classification of urban lan...SP_3 Automatic identification of high streets and classification of urban lan...
SP_3 Automatic identification of high streets and classification of urban lan...GISRUK conference
 
9B_1_Trust in web gis a preliminary investigation of the environment agencys ...
9B_1_Trust in web gis a preliminary investigation of the environment agencys ...9B_1_Trust in web gis a preliminary investigation of the environment agencys ...
9B_1_Trust in web gis a preliminary investigation of the environment agencys ...GISRUK conference
 
9A_2_Automatic classification of retail spaces from a large scale topographc ...
9A_2_Automatic classification of retail spaces from a large scale topographc ...9A_2_Automatic classification of retail spaces from a large scale topographc ...
9A_2_Automatic classification of retail spaces from a large scale topographc ...GISRUK conference
 
9A_1_On automatic mapping of environmental data using adaptive general regres...
9A_1_On automatic mapping of environmental data using adaptive general regres...9A_1_On automatic mapping of environmental data using adaptive general regres...
9A_1_On automatic mapping of environmental data using adaptive general regres...GISRUK conference
 
8B_4_Exploring the usability of geographic information
8B_4_Exploring the usability of geographic information8B_4_Exploring the usability of geographic information
8B_4_Exploring the usability of geographic informationGISRUK conference
 
8B_2_Using sound to represent uncertainty in address locations
8B_2_Using sound to represent uncertainty in address locations8B_2_Using sound to represent uncertainty in address locations
8B_2_Using sound to represent uncertainty in address locationsGISRUK conference
 
8B_1_A map to hear - use of sound in enhancing the map use experience
8B_1_A map to hear - use of sound in enhancing the map use experience8B_1_A map to hear - use of sound in enhancing the map use experience
8B_1_A map to hear - use of sound in enhancing the map use experienceGISRUK conference
 

More from GISRUK conference (20)

8A_1_To vote or not to vote
8A_1_To vote or not to vote8A_1_To vote or not to vote
8A_1_To vote or not to vote
 
7B_3_Matterhorn on the horizon
7B_3_Matterhorn on the horizon7B_3_Matterhorn on the horizon
7B_3_Matterhorn on the horizon
 
7B_2_Topological consistent generalization of openstreetmap
7B_2_Topological consistent generalization of openstreetmap7B_2_Topological consistent generalization of openstreetmap
7B_2_Topological consistent generalization of openstreetmap
 
7A_4_Gps data collection setting for pedestrian activity modelling
7A_4_Gps data collection setting for pedestrian activity modelling7A_4_Gps data collection setting for pedestrian activity modelling
7A_4_Gps data collection setting for pedestrian activity modelling
 
5A_3_GIS based spatial modelling for improving the sustainability of aggregat...
5A_3_GIS based spatial modelling for improving the sustainability of aggregat...5A_3_GIS based spatial modelling for improving the sustainability of aggregat...
5A_3_GIS based spatial modelling for improving the sustainability of aggregat...
 
5A_1_Land evaluation techniques comparing fuzzy ahp with ideal point methods
5A_1_Land evaluation techniques comparing fuzzy ahp with ideal point methods5A_1_Land evaluation techniques comparing fuzzy ahp with ideal point methods
5A_1_Land evaluation techniques comparing fuzzy ahp with ideal point methods
 
4B_1_How many volunteers does it take to map an area well
4B_1_How many volunteers does it take to map an area well4B_1_How many volunteers does it take to map an area well
4B_1_How many volunteers does it take to map an area well
 
4A_1_Uncertainty in the 2001 output area classification for the census of eng...
4A_1_Uncertainty in the 2001 output area classification for the census of eng...4A_1_Uncertainty in the 2001 output area classification for the census of eng...
4A_1_Uncertainty in the 2001 output area classification for the census of eng...
 
3A_4_Applying network analysis to quantify accessibility to urban greenspace ...
3A_4_Applying network analysis to quantify accessibility to urban greenspace ...3A_4_Applying network analysis to quantify accessibility to urban greenspace ...
3A_4_Applying network analysis to quantify accessibility to urban greenspace ...
 
3A_2_Modelling health-harming behaviours in a socially ranked geographic space
3A_2_Modelling health-harming behaviours in a socially ranked geographic space3A_2_Modelling health-harming behaviours in a socially ranked geographic space
3A_2_Modelling health-harming behaviours in a socially ranked geographic space
 
1A_3_A geodemographic classification of london primary schools
1A_3_A geodemographic classification of london primary schools1A_3_A geodemographic classification of london primary schools
1A_3_A geodemographic classification of london primary schools
 
UK Map Challenge Aidan Slingsby
UK Map Challenge   Aidan SlingsbyUK Map Challenge   Aidan Slingsby
UK Map Challenge Aidan Slingsby
 
SP_4 Supporting spatial negotiations in land use planning
SP_4 Supporting spatial negotiations in land use planningSP_4 Supporting spatial negotiations in land use planning
SP_4 Supporting spatial negotiations in land use planning
 
SP_3 Automatic identification of high streets and classification of urban lan...
SP_3 Automatic identification of high streets and classification of urban lan...SP_3 Automatic identification of high streets and classification of urban lan...
SP_3 Automatic identification of high streets and classification of urban lan...
 
9B_1_Trust in web gis a preliminary investigation of the environment agencys ...
9B_1_Trust in web gis a preliminary investigation of the environment agencys ...9B_1_Trust in web gis a preliminary investigation of the environment agencys ...
9B_1_Trust in web gis a preliminary investigation of the environment agencys ...
 
9A_2_Automatic classification of retail spaces from a large scale topographc ...
9A_2_Automatic classification of retail spaces from a large scale topographc ...9A_2_Automatic classification of retail spaces from a large scale topographc ...
9A_2_Automatic classification of retail spaces from a large scale topographc ...
 
9A_1_On automatic mapping of environmental data using adaptive general regres...
9A_1_On automatic mapping of environmental data using adaptive general regres...9A_1_On automatic mapping of environmental data using adaptive general regres...
9A_1_On automatic mapping of environmental data using adaptive general regres...
 
8B_4_Exploring the usability of geographic information
8B_4_Exploring the usability of geographic information8B_4_Exploring the usability of geographic information
8B_4_Exploring the usability of geographic information
 
8B_2_Using sound to represent uncertainty in address locations
8B_2_Using sound to represent uncertainty in address locations8B_2_Using sound to represent uncertainty in address locations
8B_2_Using sound to represent uncertainty in address locations
 
8B_1_A map to hear - use of sound in enhancing the map use experience
8B_1_A map to hear - use of sound in enhancing the map use experience8B_1_A map to hear - use of sound in enhancing the map use experience
8B_1_A map to hear - use of sound in enhancing the map use experience
 

4B_3_Automatically generating keywods for georeferenced imaged

  • 1. Automatically generating keywords for georeferenced images Ross Purves1, Alistair Edwardes1, Xin Fan2, Mark Hall2 and Martin Tomko1 1University of Zurich 2University of Sheffield 3Cardiff University
  • 2. What I want to talk about… How are images indexed? How do we describe images? How can we (semi?)-automaticallygenerate (geographically related) keywords that describe image content?
  • 3. Image indexing Image indexing is crucial, since image search is almost always based on keywords Keywords basically assigned in four ways: Manual annotation Terms freely chosen by indexers (these might include tags nowadays) Terms selected from a controlled vocabulary Automatic annotation Terms extracted from text thought to be related to an image Terms assigned to an image on the basis of content-based techniques
  • 4. …but… Image search (based on such annotation) often does not meet user expectations
  • 5. The semantic gap… “…the lack of coincidence between the information that one can extract from the visual data and the interpretation that the same data have for a user in a given situation.” (Smeulders et al., 2000)
  • 6. …in other words Search fails because annotations are not the same as search terms… …unlike text search – we are using a proxy for content The challenge is, therefore, to develop methods that better match user expectations -> and are therefore more universal Taken up by the ESP game/ Google’s image labeller
  • 7. How should we describe these pictures? Image Javier Corripio Image Google Streetview
  • 8. Theory of image description Panofsky-Shatford facet matrix – Shatford (1986)
  • 9. The where facet In the Tripod project, we are especially interested in describing images based on where they were taken We suppose, that in the medium-term all images will be georeferenced Panofsky-Shatfordmatrix suggests some ways we might describe images I’m going to concentrate on where/generic of (Martin will talk tomorrow about one element of the where/ specific of)
  • 10. Where/ generic of The generic of represents a kind of place Kinds may relate to basic levels Basic levels (e.g. Rosch, 1977) are terms used in natural language which are informative and summative (e.g. table vs. furniture or square table) Basic levels are probably very good indexing terms For example: Mountains, valleys, desert, ravines Street, pavement, house Empirical research has explored what these basic levels are in human subject experiments and we explored them in UGC/VGI How can we generate terms which relate to these kinds automatically?
  • 11. Basic process Start with an image associated with a coordinates and (sometimes) direction Basic process Identify potential visible area Query spatial data within visible area for candidate keywords Rank and filter candidate keywords to generate final keyword list
  • 12. Identifying visible area Camera parameters extracted from EXIF Content-based check for building combined with landcover data to determine urban or rural case – controls range of viewshed If no direction information, 360° viewshedgenerated, otherwise sector defined on basis of camera parameters
  • 13. Data: OpenStreetMap, ©SwissTopo and ©Ordnance Survey
  • 14. Basic keywording process Identify available data in region Query data using viewshed for data classes Map data classes to (potentially multiple) concepts – remove duplicates Concepts expanded to multiple (potentially multi-lingual) candidate keywords Rank candidate keywords according to area, probabilistic salience and web salience
  • 15.
  • 19.
  • 20.
  • 22. Returned ranked by area in viewshedWoodland, Lakes, NaturalGrassland, Reservoir, SingleBuiltWorksByFunction, Tracks, Moor, Streets, Stream, Pond, Forest, ConiferousTrees, Tarn, Heath, Byways
  • 23. Final filtering and ranking Ranking and filtering based on spatial extent, descriptive and web salience Spatial extent: concepts covering large area typically important (but favours landcover related concepts) Descriptive salience: weight rare concepts with respect to surroundings higher than common (e.g. a village shop is more salient than one on a high street) Web salience: Query web with keywordsand local toponymsto find common combinations
  • 25. Keywords: Lake, Meadows, Settlement, Reservoir, Forest
  • 26. Keywords:Street, Footpath, Gallery, Steps, Parkland, Broadway
  • 27. Are we any good overall? Initial results (with an interim solution) – about 40% of keywords are good or very good for a set of 20 images, assessed by ~70 users
  • 28. Where are we? We can (semi) automatically annotate images based on their location, with keywords related to geography Concept ontology allows us to switch data providers in and out easily Filtering and ranking reduces initial long list of candidate concepts to set of keywords ~40% of keywords are good/very good
  • 29. Some lessons learned Many devices still relatively error prone (particularly with respect to direction) Open Street Map has very rich (but sometimes esoteric) attribution – but richness in urban areas very beneficial National Mapping Agency data essential in rural areas General classes (e.g. Building) difficult to use (either too many or too general keywords)
  • 30. 25 Acknowledgements I’d like to gratefully acknowledge contributors to Geograph British Isles, see http://www.geograph.org.uk/credits/2007-02-24, whose work is made available under the following Creative Commons Attribution-ShareAlike 2.5 Licence (http://creativecommons.org/licenses/by-sa/2.5/). Much of the research reported here was part of the project TRIPOD supported by the European Commission under contract 045335.
  • 33. zurich_20070709_45.jpg Kreuzplatz photographed at 3.44 pm at the corner of Utoquai and Schoeckstrasse west of Bellevue in the Zürich (Region) Keywords: Residence, Settlement, Building, Byway, Lane, Street, Lake, Road, Pool Datasets: SwisstopoVec 25, OSM Spotted feature types: Vec25: 4_klass, 1_klass, Str_bahn, Z_innenhof, Gedbrue, Z_gebaeude, Z_station, Z_kirche, Ww50, Z_bharea, Z_see, Obstbaum, See, 2_klass, Obreihe, Kiturm, Z_siedl OSM: Living_street, Rail_station_poly, Service, Wiese, Museum, Pedestrian, Cycleway, Rail, Retail, Pear_a, Theatre, University_complet, Primary, Residential, Footway, Secundary, Path, Water, School_a, Spielplatz, Steps, Place_of_worship, Cemetery, Buildings, Platform, Tertiary, Wohngebiet, Concepts: Trees, Turnpikes, SingleBuiltWorksByFunction, Towpaths, Streets, Footpaths, Cathedral, MotorwayFlyover, Byways, Waterbodies, Bridges, Trains, Steps, Motorways, Roads, MotorwayExtension, Museums, Reservoir, Lanes, ChurchTower, Bypasses, Meadow, RailwayTracks, Highways, Trackbed, Theatres, Chapels, Graveyard, Orchards, Churchyard, Colleges, Pond, Housing, CyclePaths, Tramway, Mainline, Broadways, Churches, Shops, Temples, Lakes, Abbey, RailwayStations, RailwaySidings, Platforms, ReligiousStructures, Courtyard, Cemeteries, Settlements, Paths, Greenways, Tarn, University, Playground, Schools,
  • 34.
  • 35. Keywords:Meadow Downs Fields Beech Wood Villages
  • 36. Datasets: Corine, OS Strategi, OS VectorMap
  • 40. VectorMap:PrivateRoadRestrictedAlignment, Orchard, MinorRoadAlignment, BroadleafedWoodlandAndShrub, UnimprovedGrass, ImportantBuilding, Marsh, ConiferousWoodland, LocalStreetAlignment, WaterBodies, Shrub, MixedWoodland, BRoadAlignment, MixedWoodlandAndShrub, UrbanExtent, Heathland, BroadleafedWoodland, WaterFeature, BuildingPolygon, ARoadPrimaryAlignment, ShrubAndUnimprovedGrass, GlasshousePolygon
  • 42. Hamlets, NaturalGrassland, Undergrowth, SingleBuiltWorksByFunction, Villages, Streets, Orchards, Bushes, AgriculturalLand, Heath, Dairy, Byways, ShrubAssociations, Copse, TransitionalWoodland, Woodland, DeciduousTrees, Marshes, Hedgerows, City, Pasture, Meadow, Settlements, Stream, Forest, Towns, Grass