Repurposing a deep learning network to filter and classify volunteered photographs for land cover and land use characterization
This paper extends recent research into the usefulness of volunteered photos for land cover extraction, and investigates whether this usefulness can be automatically assessed by an easily accessible, off-the-shelf neural network pre-trained on a variety of scene characteristics. Geo-tagged photograp...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2017-07-01
|
| Series: | Geo-spatial Information Science |
| Subjects: | |
| Online Access: | http://dx.doi.org/10.1080/10095020.2017.1373955 |
| _version_ | 1818481316313694208 |
|---|---|
| author | Lukasz Tracewski Lucy Bastin Cidalia C. Fonte |
| author_facet | Lukasz Tracewski Lucy Bastin Cidalia C. Fonte |
| author_sort | Lukasz Tracewski |
| collection | DOAJ |
| description | This paper extends recent research into the usefulness of volunteered photos for land cover extraction, and investigates whether this usefulness can be automatically assessed by an easily accessible, off-the-shelf neural network pre-trained on a variety of scene characteristics. Geo-tagged photographs are sometimes presented to volunteers as part of a game which requires them to extract relevant facts about land use. The challenge is to select the most relevant photographs in order to most efficiently extract the useful information while maintaining the engagement and interests of volunteers. By repurposing an existing network which had been trained on an extensive library of potentially relevant features, we can quickly carry out initial assessments of the general value of this approach, pick out especially salient features, and identify focus areas for future neural network training and development. We compare two approaches to extract land cover information from the network: a simple post hoc weighting approach accessible to non-technical audiences and a more complex decision tree approach that involves training on domain-specific features of interest. Both approaches had reasonable success in characterizing human influence within a scene when identifying the land use types (as classified by Urban Atlas) present within a buffer around the photograph’s location. This work identifies important limitations and opportunities for using volunteered photographs as follows: (1) the false precision of a photograph’s location is less useful for identifying on-the-spot land cover than the information it can give on neighbouring combinations of land cover; (2) ground-acquired photographs, interpreted by a neural network, can supplement plan view imagery by identifying features which will never be discernible from above; (3) when dealing with contexts where there are very few exemplars of particular classes, an independent a posteriori weighting of existing scene attributes and categories can buffer against over-specificity. |
| first_indexed | 2024-12-10T11:33:17Z |
| format | Article |
| id | doaj.art-3744fc8210764fceb1a2d496c3d3bda4 |
| institution | Directory Open Access Journal |
| issn | 1009-5020 1993-5153 |
| language | English |
| last_indexed | 2024-12-10T11:33:17Z |
| publishDate | 2017-07-01 |
| publisher | Taylor & Francis Group |
| record_format | Article |
| series | Geo-spatial Information Science |
| spelling | doaj.art-3744fc8210764fceb1a2d496c3d3bda42022-12-22T01:50:29ZengTaylor & Francis GroupGeo-spatial Information Science1009-50201993-51532017-07-0120325226810.1080/10095020.2017.13739551373955Repurposing a deep learning network to filter and classify volunteered photographs for land cover and land use characterizationLukasz Tracewski0Lucy Bastin1Cidalia C. Fonte2Aston UniversityAston UniversityINESC Coimbra, University of CoimbraThis paper extends recent research into the usefulness of volunteered photos for land cover extraction, and investigates whether this usefulness can be automatically assessed by an easily accessible, off-the-shelf neural network pre-trained on a variety of scene characteristics. Geo-tagged photographs are sometimes presented to volunteers as part of a game which requires them to extract relevant facts about land use. The challenge is to select the most relevant photographs in order to most efficiently extract the useful information while maintaining the engagement and interests of volunteers. By repurposing an existing network which had been trained on an extensive library of potentially relevant features, we can quickly carry out initial assessments of the general value of this approach, pick out especially salient features, and identify focus areas for future neural network training and development. We compare two approaches to extract land cover information from the network: a simple post hoc weighting approach accessible to non-technical audiences and a more complex decision tree approach that involves training on domain-specific features of interest. Both approaches had reasonable success in characterizing human influence within a scene when identifying the land use types (as classified by Urban Atlas) present within a buffer around the photograph’s location. This work identifies important limitations and opportunities for using volunteered photographs as follows: (1) the false precision of a photograph’s location is less useful for identifying on-the-spot land cover than the information it can give on neighbouring combinations of land cover; (2) ground-acquired photographs, interpreted by a neural network, can supplement plan view imagery by identifying features which will never be discernible from above; (3) when dealing with contexts where there are very few exemplars of particular classes, an independent a posteriori weighting of existing scene attributes and categories can buffer against over-specificity.http://dx.doi.org/10.1080/10095020.2017.1373955Land coverland usevolunteered geographic information (VGI)photographconvolutional neural networkmachine learning |
| spellingShingle | Lukasz Tracewski Lucy Bastin Cidalia C. Fonte Repurposing a deep learning network to filter and classify volunteered photographs for land cover and land use characterization Geo-spatial Information Science Land cover land use volunteered geographic information (VGI) photograph convolutional neural network machine learning |
| title | Repurposing a deep learning network to filter and classify volunteered photographs for land cover and land use characterization |
| title_full | Repurposing a deep learning network to filter and classify volunteered photographs for land cover and land use characterization |
| title_fullStr | Repurposing a deep learning network to filter and classify volunteered photographs for land cover and land use characterization |
| title_full_unstemmed | Repurposing a deep learning network to filter and classify volunteered photographs for land cover and land use characterization |
| title_short | Repurposing a deep learning network to filter and classify volunteered photographs for land cover and land use characterization |
| title_sort | repurposing a deep learning network to filter and classify volunteered photographs for land cover and land use characterization |
| topic | Land cover land use volunteered geographic information (VGI) photograph convolutional neural network machine learning |
| url | http://dx.doi.org/10.1080/10095020.2017.1373955 |
| work_keys_str_mv | AT lukasztracewski repurposingadeeplearningnetworktofilterandclassifyvolunteeredphotographsforlandcoverandlandusecharacterization AT lucybastin repurposingadeeplearningnetworktofilterandclassifyvolunteeredphotographsforlandcoverandlandusecharacterization AT cidaliacfonte repurposingadeeplearningnetworktofilterandclassifyvolunteeredphotographsforlandcoverandlandusecharacterization |