{"id":{"repo_id":"malta","oai_identifier":"oai:www.um.edu.mt:123456789/25487"},"canonical_url":"https://search.dev.ndltd.org/etd/malta/oai:www.um.edu.mt:123456789/25487","repository":{"repo_id":"malta","name":"University of Malta","base_url":"https://www.um.edu.mt/library/oar/oai/request"},"display":{"title":"Automatic segmentation of indoor and outdoor scenes from visual lifelogging","abstract":"Visual lifelogging and indoor & outdoor classification have been researched on several occasions. The focus of this research is to identify and implement the most effective methods to automatically classify images, captured from a wearable camera, into indoor and outdoor scenes. We incorporated these statistics in a smart phone application where the user can monitor his/her activities. To automatically classify images into indoor & outdoor scenes, we have used pre-trained convolutional neural networks on Imagenet and Places2 and applied transfer learning to adapt to the problem at hand. This method proved to be useful instantly, obtaining accuracies above 90%. Therefore, we investigated ways to improve the classification further by joining different features, using intermediate layer information from the CNNs and improving the classifiers used. When tested on the UBRug dataset, our best method, joining the FC7 intermediate layer's output from both CNNs with a Random Forest classifier, obtained around 98% accuracy. We compared our results with previous tests on the SUN397 dataset, our methods outperformed these tests. The best performing method from these tests obtained an accuracy rate of 94.2% while our best performing method obtained an accuracy of 97.06%. Moreover, we tested re-training a pre-trained CNN on the UBRug Dataset, introducing, the first Ego-centric indoor/outdoor CNN.","abstract_html":"Visual lifelogging and indoor & outdoor classification have been researched on several occasions. The focus of this research is to identify and implement the most effective methods to automatically classify images, captured from a wearable camera, into indoor and outdoor scenes. We incorporated these statistics in a smart phone application where the user can monitor his/her activities. To automatically classify images into indoor & outdoor scenes, we have used pre-trained convolutional neural networks on Imagenet and Places2 and applied transfer learning to adapt to the problem at hand. This method proved to be useful instantly, obtaining accuracies above 90%. Therefore, we investigated ways to improve the classification further by joining different features, using intermediate layer information from the CNNs and improving the classifiers used. When tested on the UBRug dataset, our best method, joining the FC7 intermediate layer's output from both CNNs with a Random Forest classifier, obtained around 98% accuracy. We compared our results with previous tests on the SUN397 dataset, our methods outperformed these tests. The best performing method from these tests obtained an accuracy rate of 94.2% while our best performing method obtained an accuracy of 97.06%. Moreover, we tested re-training a pre-trained CNN on the UBRug Dataset, introducing, the first Ego-centric indoor/outdoor CNN.","abstract_has_math":false,"creators":[],"institution":"University of Malta","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2017,"date_issued":"2017","date_published":"2017","updated_at":"2026-07-27T20:11:20Z","subjects":["Data logging","Wearable cameras","Neural networks (Computer science)","Computer vision"],"languages":["en"],"rights":["info:eu-repo/semantics/restrictedAccess"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://www.um.edu.mt/library/oar//handle/123456789/25487","outbound_label":"Repository record","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2018-01-05T15:30:25Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2018-01-05T15:30:25Z"]},{"key":"dc:date.issued","label":"Date","values":["2017"]},{"key":"dc:publisher.department","label":"Dc Publisher Department","values":["Faculty of Information and Communication Technology. Department of Artificial Intelligence"]},{"key":"dc:publisher.institution","label":"Dc Publisher Institution","values":["University of Malta"]},{"key":"dc:type","label":"Dc Type","values":["bachelorThesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Data logging","Wearable cameras","Neural networks (Computer science)","Computer vision"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["info:eu-repo/semantics/restrictedAccess"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://www.um.edu.mt/library/oar//handle/123456789/25487"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["B.SC.(HONS)ICT"]},{"key":"dc:description.abstract","label":"Abstract","values":["Visual lifelogging and indoor & outdoor classification have been researched on several occasions. The focus of this research is to identify and implement the most effective methods to automatically classify images, captured from a wearable camera, into indoor and outdoor scenes. We incorporated these statistics in a smart phone application where the user can monitor his/her activities. To automatically classify images into indoor & outdoor scenes, we have used pre-trained convolutional neural networks on Imagenet and Places2 and applied transfer learning to adapt to the problem at hand. This method proved to be useful instantly, obtaining accuracies above 90%. Therefore, we investigated ways to improve the classification further by joining different features, using intermediate layer information from the CNNs and improving the classifiers used. When tested on the UBRug dataset, our best method, joining the FC7 intermediate layer's output from both CNNs with a Random Forest classifier, obtained around 98% accuracy. We compared our results with previous tests on the SUN397 dataset, our methods outperformed these tests. The best performing method from these tests obtained an accuracy rate of 94.2% while our best performing method obtained an accuracy of 97.06%. Moreover, we tested re-training a pre-trained CNN on the UBRug Dataset, introducing, the first Ego-centric indoor/outdoor CNN."]},{"key":"dc:title","label":"Title","values":["Automatic segmentation of indoor and outdoor scenes from visual lifelogging"]}]}],"canonical_facts":{"dc:date.accessioned":["2018-01-05T15:30:25Z"],"dc:date.available":["2018-01-05T15:30:25Z"],"dc:date.issued":["2017"],"dc:description":["B.SC.(HONS)ICT"],"dc:description.abstract":["Visual lifelogging and indoor & outdoor classification have been researched on several occasions. The focus of this research is to identify and implement the most effective methods to automatically classify images, captured from a wearable camera, into indoor and outdoor scenes. We incorporated these statistics in a smart phone application where the user can monitor his/her activities. To automatically classify images into indoor & outdoor scenes, we have used pre-trained convolutional neural networks on Imagenet and Places2 and applied transfer learning to adapt to the problem at hand. This method proved to be useful instantly, obtaining accuracies above 90%. Therefore, we investigated ways to improve the classification further by joining different features, using intermediate layer information from the CNNs and improving the classifiers used. When tested on the UBRug dataset, our best method, joining the FC7 intermediate layer's output from both CNNs with a Random Forest classifier, obtained around 98% accuracy. We compared our results with previous tests on the SUN397 dataset, our methods outperformed these tests. The best performing method from these tests obtained an accuracy rate of 94.2% while our best performing method obtained an accuracy of 97.06%. Moreover, we tested re-training a pre-trained CNN on the UBRug Dataset, introducing, the first Ego-centric indoor/outdoor CNN."],"dc:identifier.uri":["https://www.um.edu.mt/library/oar//handle/123456789/25487"],"dc:language.iso":["en"],"dc:publisher.department":["Faculty of Information and Communication Technology. Department of Artificial Intelligence"],"dc:publisher.institution":["University of Malta"],"dc:rights":["info:eu-repo/semantics/restrictedAccess"],"dc:subject":["Data logging","Wearable cameras","Neural networks (Computer science)","Computer vision"],"dc:title":["Automatic segmentation of indoor and outdoor scenes from visual lifelogging"],"dc:type":["bachelorThesis"]},"updated_at":"2026-07-27T20:11:20Z"}